A chronic disease drug monitoring method and system based on digital twinning

Abstract

The application discloses a chronic disease drug monitoring method and system based on digital twinning, comprising the following steps: acquiring and preprocessing the course data of a target patient; constructing a patient digital twinning course body; extracting the current course state parameters and constructing a stable control domain; constructing a virtual drug disturbance condition set; loading each virtual drug disturbance condition to the patient digital twinning course body, performing parallel course evolution deduction and generating multiple future course evolution paths; determining the bifurcation determination parameters and the virtual course bifurcation points corresponding to the multiple future course evolution paths; determining the sensitive drug factors and generating the chronic disease drug pre-monitoring result. The application uses the digital twinning course deduction method to realize the pre-monitoring of the chronic disease drug risk, and has the advantages of strong predictability, high monitoring accuracy and timely intervention.

Description

Technical Field

[0001] This invention relates to the field of chronic disease health management, and in particular to a method and system for monitoring medication use in chronic diseases based on digital twins. Background Technology

[0002] With the development of digital healthcare, wearable devices, and remote health monitoring technologies, the long-term medication management of patients with chronic diseases is gradually shifting from traditional manual follow-up to intelligent monitoring. Existing technologies typically analyze patients' current health status by collecting data such as blood pressure, blood sugar, heart rate, and historical medication records, and then combine this with rule-based or statistical models to generate medication reminders and risk warnings, thereby assisting doctors and patients in managing chronic diseases.

[0003] Most existing technologies rely on static indicators or single monitoring results for risk assessment, lacking the ability to dynamically extrapolate the continuous evolution of a patient's disease course. This makes it difficult to predict future disease progression by incorporating different medication-related disturbances. Furthermore, current solutions lack synergistic analysis of complex confounding factors such as diet, exercise, and combined medication use, failing to identify sensitive medication factors that exacerbate disease progression. This results in a lag in chronic disease risk warnings, making it difficult to meet the needs for proactive and refined medication monitoring in complex chronic disease scenarios.

[0004] Therefore, how to provide a method and system for monitoring chronic disease medication based on digital twins is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] One objective of this invention is to propose a method and system for monitoring medication use in chronic diseases based on digital twins. This invention utilizes the digital twin disease progression extrapolation method to achieve proactive monitoring of medication risks in chronic diseases, and has the advantages of strong predictability, high monitoring accuracy, and timely intervention.

[0006] A method for monitoring medication use in chronic diseases based on digital twins according to an embodiment of the present invention includes the following steps:

[0007] Acquire the patient's medical history data, preprocess it, and generate a fusion dataset of the patient's medical history.

[0008] Constructing digital twin disease history data of patients based on a fusion dataset of patient disease history data;

[0009] Extract the current disease status parameters from the patient's disease course fusion dataset, map the current disease status parameters to the patient's digital twin disease course to generate the current disease status vector, and construct a stable control domain by combining the disease status distribution corresponding to the patient's historical effective control stages.

[0010] Based on the target patient’s current medication regimen, historical abnormal medication behavior, and records of lifestyle disturbances, a set of virtual medication disturbance conditions is constructed.

[0011] Under the same current disease state vector, each virtual medication perturbation condition is loaded into the patient's digital twin disease process, parallel disease evolution inference is performed, and multiple future disease evolution paths are generated;

[0012] Path separation analysis, stability control domain deviation analysis, and risk branch generation analysis are performed on multiple future disease evolution paths to obtain bifurcation determination parameters, and virtual disease bifurcation points corresponding to multiple future disease evolution paths are determined based on the bifurcation determination parameters.

[0013] Based on the virtual disease course bifurcation point, the sensitive medication factors that trigger disease course bifurcation are identified, and the pre-monitoring results of chronic disease medication are generated.

[0014] Optionally, the disease course data includes basic information on chronic diseases, historical diagnosis and treatment information, historical medication information, continuous physiological monitoring information, lifestyle behavior information, and disease assessment information. The data preprocessing includes time alignment, missing data completion, indicator normalization, abnormal data removal, and multi-source association mapping.

[0015] Optionally, the construction of the patient's digital twin pathology specifically includes:

[0016] Based on the basic information on chronic diseases and historical diagnosis and treatment information in the patient disease course fusion dataset, a disease status mapping unit is established;

[0017] Based on the historical medication information and current medication regimen in the patient's disease course fusion dataset, a medication intervention mapping unit is established;

[0018] A physiological regulation mapping unit is established based on continuous physiological monitoring information in the patient's disease course fusion dataset;

[0019] Based on the lifestyle behavior information in the patient's disease course fusion dataset, a lifestyle behavior disturbance mapping unit is established.

[0020] Based on the mapping relationships between the disease state mapping unit, the medication intervention mapping unit, the physiological regulation mapping unit, and the lifestyle behavior disturbance mapping unit, a disease course evolution simulation unit is established.

[0021] By associating and encapsulating the disease state mapping unit, medication intervention mapping unit, physiological regulation mapping unit, lifestyle behavior disturbance mapping unit, and disease course evolution simulation unit, a digital twin disease course of the patient is constructed.

[0022] Optionally, the construction of the stable control domain specifically includes:

[0023] Extract the disease course data corresponding to the current time window of the target patient from the patient disease course fusion dataset, and extract the current disease status parameters, current medication status parameters, current physiological regulation parameters and current lifestyle behavior disturbance parameters based on the disease course data corresponding to the current time window to form the current disease course status parameters;

[0024] Input the current disease state parameters into the disease state mapping unit, the current medication state parameters into the medication intervention mapping unit, the current physiological regulation parameters into the physiological regulation mapping unit, and the current lifestyle disturbance parameters into the lifestyle disturbance mapping unit to obtain the disease state mapping value, medication intervention mapping value, physiological regulation mapping value, and lifestyle disturbance mapping value, respectively.

[0025] Based on the state associations in the patient's digital twin disease course, the disease state mapping value, medication intervention mapping value, physiological regulation mapping value, and lifestyle behavior disturbance mapping value are spliced, weighted, and time-series labeled to generate the current disease course state vector.

[0026] Extract disease status samples corresponding to the historical effective control stages of patients from the patient disease course fusion dataset, perform steady-state distribution modeling on the disease status samples, and obtain the disease status distribution corresponding to the historical effective control stages of patients.

[0027] The stable state boundary is determined based on the distribution of the disease course state, and the current disease course state vector is mapped to the state space corresponding to the range of the stable state boundary to construct a stable control domain.

[0028] Optionally, the construction of the virtual medication perturbation condition set specifically includes:

[0029] Extract drug names, dosages, administration times, administration frequencies, and combination drug relationships from the target patient's current medication regimen to construct normal medication conditions;

[0030] Based on the records of delayed medication, missed medication, make-up medication, insufficient dosage, and excessive dosage in historical abnormal medication behavior, conditions for delayed medication, missed medication, make-up medication, insufficient dosage, and excessive dosage are constructed respectively.

[0031] Based on dietary and exercise records in the life behavior disturbance records, dietary disturbance conditions and exercise disturbance conditions were constructed respectively.

[0032] Based on the combination drug relationships in the current medication regimen and the abnormal combination drug records in historical abnormal medication behaviors, construct the combination drug interference conditions;

[0033] By labeling the perturbation type and the deduction order of normal medication conditions, delayed medication conditions, missed medication conditions, make-up medication conditions, insufficient dosage conditions, excessive dosage conditions, dietary interference conditions, exercise interference conditions, and combined medication interference conditions, a set of virtual medication perturbation conditions is constructed.

[0034] Optionally, the generation of the multiple future disease progression paths specifically includes:

[0035] Extract each virtual medication perturbation condition and its loading order from the set of virtual medication perturbation conditions, take the current disease state vector as the common initial state, and establish corresponding parallel deduction channels according to each virtual medication perturbation condition.

[0036] According to the loading order, each virtual medication disturbance condition is loaded into the medication intervention mapping unit and the life behavior disturbance mapping unit in the patient's digital twin disease process;

[0037] Within the same simulation time window, the disease progression simulation unit performs disease progression simulation on each parallel simulation channel, and updates the disease state mapping value, medication intervention mapping value, physiological regulation mapping value and lifestyle disturbance mapping value in sequence according to the simulation time nodes.

[0038] For each parallel simulation channel, the disease state mapping value, medication intervention mapping value, physiological regulation mapping value, and lifestyle disturbance mapping value corresponding to each simulation time node are associated and arranged in chronological order to generate the disease state evolution sequence under the corresponding virtual medication disturbance conditions.

[0039] Each disease progression sequence is labeled and aggregated according to the corresponding virtual medication perturbation conditions to generate multiple future disease progression paths.

[0040] Optionally, determining the virtual disease course bifurcation point specifically includes:

[0041] Multiple future disease progression paths are time-aligned according to the same projection time node. The disease state mapping value, physiological regulation mapping value, and disease stage change state corresponding to each future disease progression path at each projection time node are extracted to form a path node state set.

[0042] Path separation analysis is performed on the set of path node states. The differences in disease state mapping values, physiological regulation mapping values, and disease stage levels between different future disease evolution paths at the same projection time node are calculated and weighted to obtain the path dispersion.

[0043] A stable control domain deviation analysis is performed on the set of path node states. The path node states corresponding to each simulation time node are mapped to the stable control domain. The boundary distance, out-of-bounds time length, and out-of-bounds state intensity between the path node states and the stable control domain boundary are determined and weighted to obtain the stable domain deviation.

[0044] Risk branch generation analysis is performed on the set of path node states to identify risk paths that transition from a stable control phase or a volatility control phase to a risk deterioration phase. The risk branch generation intensity is obtained by weighted summarization based on the number of risk paths, the duration of risk paths, and the magnitude of risk level transitions.

[0045] The path dispersion, stability domain deviation, and risk branch generation intensity are used as bifurcation determination parameters. The bifurcation determination parameters are continuously determined according to the order of the simulation time nodes, and the simulation time node that first continuously meets the corresponding preset determination conditions is determined as the virtual disease course bifurcation point.

[0046] Optionally, the generation of the pre-monitoring results for chronic disease medication specifically includes:

[0047] Extract the simulation time nodes, bifurcation paths, and virtual medication disturbance conditions corresponding to the bifurcation points in the virtual disease course, and determine the bifurcation occurrence time window based on the simulation time nodes;

[0048] Extract the disease state mapping value, physiological regulation mapping value and disease stage change status corresponding to the continuous extrapolation time nodes before the bifurcation time window from the bifurcation path to generate disease state change characteristics before bifurcation.

[0049] The characteristics of disease course status changes before bifurcation are correlated with the virtual medication perturbation conditions corresponding to the virtual disease course bifurcation points to determine the path dispersion changes, stability domain deviation changes, and risk branch generation intensity changes corresponding to each virtual medication perturbation condition.

[0050] Based on the changes in path dispersion, stability domain deviation, and risk branch generation intensity corresponding to each virtual medication perturbation condition, the virtual medication perturbation conditions that lead to the bifurcation of future disease evolution paths are identified, and the identified virtual medication perturbation conditions are determined as sensitive medication factors.

[0051] The target patient identifier is extracted from the patient disease course fusion dataset, the chronic disease type is extracted from the disease status mapping results, the risk branch type is determined according to the disease stage change status corresponding to the virtual disease course bifurcation point, the medication monitoring prompt is generated according to the sensitive medication factors, the bifurcation occurrence time window and the risk branch type, and the target patient identifier, chronic disease type, sensitive medication factors, bifurcation occurrence time window, risk branch type and medication monitoring prompt are combined to generate the pre-monitoring results of chronic disease medication.

[0052] A chronic disease medication monitoring system based on digital twin according to an embodiment of the present invention includes:

[0053] The data fusion module is used to acquire and preprocess the patient's medical history data.

[0054] The twin building module is used to construct digital twin disease profiles of patients based on a fusion dataset of patient disease histories.

[0055] The control domain module is used to generate the current disease state vector and construct a stable control domain.

[0056] The perturbation construction module is used to construct a set of virtual medication perturbation conditions;

[0057] The parallel simulation module is used to load various virtual medication perturbation conditions into the patient's digital twin disease process, perform parallel disease evolution simulation, and generate multiple future disease evolution paths;

[0058] The bifurcation identification module is used to obtain bifurcation determination parameters and determine virtual disease bifurcation points corresponding to multiple future disease evolution paths;

[0059] The results generation module is used to identify sensitive medication factors and generate pre-monitoring results for chronic disease medication.

[0060] The beneficial effects of this invention are:

[0061] This invention constructs a digital twin disease process model for patients, unifying and dynamically mapping disease status, medication intervention status, physiological regulation status, and lifestyle disturbance status. This enables the digital reconstruction and continuous simulation of the disease progression process in chronic disease patients. Compared to traditional methods that rely solely on single monitoring results for risk assessment, this invention combines historical disease data with the current disease status to perform parallel extrapolation of future disease progression under different medication behaviors and lifestyle disturbances. This improves the dynamic perception of chronic disease development trends, transforming chronic disease medication monitoring from a reactive response to a proactive prediction.

[0062] This invention constructs a stable control domain and performs path separation analysis, stable control domain deviation analysis, and risk branch generation analysis on multiple future disease progression paths. This allows for the identification of key bifurcation nodes that allow a patient's disease progression to shift from a stable state to a state of worsening risk. By comparing future disease progression paths corresponding to different virtual medication perturbation conditions, it can accurately pinpoint sensitive medication factors leading to disease progression, including delayed medication, missed doses, abnormal dosage, interference from combined medications, and complex influencing factors such as diet and exercise. This improves the accuracy and lead time of chronic disease risk identification, reducing the lag in risk detection in traditional monitoring methods.

[0063] This invention can also generate medication monitoring prompts based on the characteristics of disease status changes and sensitive medication factors corresponding to virtual disease bifurcation points, providing targeted pre-emptive intervention basis for patients and medical personnel. By performing correlation analysis on the risk occurrence time window, the direction of disease progression, and corresponding abnormal medication users, personalized medication monitoring results can be output in advance, enhancing the ability of refined intervention and continuous health management in the process of chronic disease management, improving patient medication adherence and the stability of disease control, and has high clinical application value and intelligent medical promotion value. Attached Figure Description

[0064] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0065] Figure 1 This is a flowchart of a chronic disease medication monitoring method based on digital twins proposed in this invention;

[0066] Figure 2 This is a flowchart illustrating the disease progression of a chronic disease medication monitoring method based on digital twins proposed in this invention.

[0067] Figure 3 This is a flowchart illustrating the virtual disease course bifurcation point determination process for a chronic disease medication monitoring method based on digital twins proposed in this invention. Detailed Implementation

[0068] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0069] refer to Figures 1-3 A method for monitoring medication use in chronic diseases based on digital twins includes the following steps:

[0070] Acquire the patient's medical history data, preprocess it, and generate a fusion dataset of the patient's medical history.

[0071] Constructing digital twin disease history data of patients based on a fusion dataset of patient disease history data;

[0072] Extract the current disease status parameters from the patient's disease course fusion dataset, map the current disease status parameters to the patient's digital twin disease course to generate the current disease status vector, and construct a stable control domain by combining the disease status distribution corresponding to the patient's historical effective control stages.

[0073] Based on the target patient’s current medication regimen, historical abnormal medication behavior, and records of lifestyle disturbances, a set of virtual medication disturbance conditions is constructed.

[0074] Under the same current disease state vector, each virtual medication perturbation condition is loaded into the patient's digital twin disease process, parallel disease evolution inference is performed, and multiple future disease evolution paths are generated;

[0075] Path separation analysis, stability control domain deviation analysis, and risk branch generation analysis are performed on multiple future disease evolution paths to obtain bifurcation determination parameters, and virtual disease bifurcation points corresponding to multiple future disease evolution paths are determined based on the bifurcation determination parameters.

[0076] Based on the virtual disease course bifurcation point, the sensitive medication factors that trigger disease course bifurcation are identified, and the pre-monitoring results of chronic disease medication are generated.

[0077] In this embodiment, the disease course data includes basic information on chronic diseases, historical diagnosis and treatment information, historical medication information, continuous physiological monitoring information, lifestyle behavior information, and disease assessment information. Data preprocessing includes time alignment, missing data completion, indicator normalization, abnormal data removal, and multi-source association mapping. The multi-source association mapping includes unifying the attribution of basic information on chronic diseases, historical diagnosis and treatment information, historical medication information, continuous physiological monitoring information, lifestyle behavior information, and disease assessment information based on patient identifiers; mapping medication time, monitoring time, behavior occurrence time, and disease assessment time based on time windows; and dividing and merging the corresponding multi-source data according to disease course stages to generate a patient disease course fusion dataset.

[0078] In this embodiment, the construction of the patient's digital twin disease profile specifically includes:

[0079] Based on the basic chronic disease information and historical treatment information in the patient disease course fusion dataset, a disease state mapping unit is established. Specifically, this involves: extracting the chronic disease name, diagnosis code, and diagnosis record from the patient disease course fusion dataset, and determining the chronic disease type corresponding to the target patient; extracting treatment time, follow-up visit records, examination results, and disease assessment records from the target patient's chronic disease type, historical treatment information, and disease assessment information in the patient disease course fusion dataset, and dividing the target patient's disease course into stages, including stable control stage, fluctuating control stage, and risk deterioration stage; and extracting disease indicators within each disease course stage according to different time windows. The system records disease indicator changes. It determines the direction of disease change based on the direction of change in disease indicator records within adjacent time windows, the magnitude of disease change based on the numerical differences in disease indicator records within adjacent time windows, the duration of disease condition based on the duration of the same disease stage, and the frequency of disease fluctuation based on the number of times disease indicator changes occur within consecutive time windows. Based on the target patient's chronic disease type, disease stage, direction of disease change, magnitude of disease change, duration of disease condition, and frequency of disease fluctuation, it establishes a correspondence between historical disease status changes and the current disease status. Based on this correspondence, it generates a disease status mapping result for the target patient and constructs a disease status mapping unit.

[0080] Based on historical medication information and current medication regimens in the patient's disease course fusion dataset, a medication intervention mapping unit is established. Specifically, this involves: extracting drug name, drug category, dosage, administration time, administration frequency, and duration of medication from the historical medication information and current medication regimens in the patient's disease course fusion dataset; extracting the target patient's actual medication records and medication regimens within each time window to form medication status data; determining the duration of drug action based on drug name, drug category, and duration of medication; determining the dosing interval change status based on the difference in administration time between the actual medication record and the medication regimen; determining the dose change status based on the difference in dosage between the actual dosage and the medication regimen; and determining the medication adherence change status based on the difference in frequency between the actual number of medications and the medication regimen; establishing a correspondence between medication intervention behavior and changes in disease course status based on the duration of drug action, dosing interval change status, dosage change status, and medication adherence change status; generating medication intervention mapping results based on the correspondence; and constructing the medication intervention mapping unit.

[0081] Based on continuous physiological monitoring information from the patient's disease course fusion dataset, a physiological regulation mapping unit is established. Specifically, this involves: extracting blood pressure, blood glucose, heart rate, blood oxygen, and body temperature indicators for the target patient within different time windows, and classifying them according to physiological indicator type to form physiological indicator classification data; arranging the physiological indicator classification data in a time series according to the collection time; determining the direction of physiological indicator change based on the rise and fall of physiological indicator values ​​within adjacent time windows; determining the amplitude of physiological indicator change based on the difference between physiological indicator values ​​within adjacent time windows; determining the physiological indicator fluctuation cycle based on the interval of repeated rise and fall changes of physiological indicators within continuous time windows; and determining the physiological indicator recovery state based on the time required for abnormal physiological indicators to return to the corresponding value range of the target patient's historical effective control stage; establishing a correspondence between physiological regulation state and disease course state changes based on the direction of physiological indicator change, physiological indicator amplitude, physiological indicator fluctuation cycle, and physiological indicator recovery state; generating physiological regulation mapping results based on the correspondence; and constructing a physiological regulation mapping unit.

[0082] Based on the lifestyle behavior information in the patient's disease course fusion dataset, a lifestyle behavior disturbance mapping unit is established. Specifically, this involves: extracting dietary records, exercise records, sleep records, and mood records from the patient's disease course fusion dataset and determining the lifestyle behavior type corresponding to the target patient; extracting the behavior occurrence time, duration, and characteristics corresponding to each lifestyle behavior type according to different time windows to form lifestyle behavior status data; determining dietary fluctuation status based on intake time, intake amount, and intake type in dietary records; determining exercise fluctuation status based on exercise time, exercise duration, and exercise intensity in exercise records; determining sleep fluctuation status based on sleep onset time, sleep duration, and number of sleep interruptions in sleep records; and determining mood fluctuation status based on mood type, mood duration, and number of mood changes in mood records; establishing a correspondence between lifestyle behavior disturbance status and disease course status changes based on dietary fluctuation status, exercise fluctuation status, sleep fluctuation status, and mood fluctuation status; generating lifestyle behavior disturbance mapping results based on the correspondence; and constructing a lifestyle behavior disturbance mapping unit.

[0083] Based on the mapping relationships between the disease state mapping unit, medication intervention mapping unit, physiological regulation mapping unit, and lifestyle behavior disturbance mapping unit, a disease progression simulation unit is established. Specifically, the disease state mapping results, medication intervention mapping results, physiological regulation mapping results, and lifestyle behavior disturbance mapping results are processed synchronously within the same time window to form a set of multiple mapping results under the same time window. Furthermore, based on the correspondences between the disease state mapping results and medication intervention mapping results, the disease state mapping results and physiological regulation mapping results, and the disease state mapping results and lifestyle behavior disturbance mapping results... State association processing is performed on the multi-mapping result set to obtain disease state association data; the trend of disease state change is determined based on the continuity of the change direction and the duration of change of the disease state association data within a continuous time window; the direction of disease state change is determined based on the change direction of the disease state association data within a continuous time window; and the intensity of disease state fluctuation is determined based on the number of fluctuations and the amplitude of fluctuations of the disease state association data within a continuous time window; an evolutionary relationship between the disease state association data and future disease state changes is established based on the disease state change trend, the change direction of disease state, and the intensity of disease state fluctuation; and a disease evolution simulation unit is constructed based on the evolutionary relationship.

[0084] By associating and encapsulating the disease state mapping unit, medication intervention mapping unit, physiological regulation mapping unit, lifestyle behavior disturbance mapping unit, and disease course evolution simulation unit, a digital twin disease course of the patient is constructed.

[0085] In this embodiment, the construction of the stable control domain specifically includes:

[0086] Extract the disease course data corresponding to the current time window of the target patient from the patient disease course fusion dataset, and extract the current disease status parameters, current medication status parameters, current physiological regulation parameters and current lifestyle behavior disturbance parameters based on the disease course data corresponding to the current time window to form the current disease course status parameters;

[0087] Input the current disease state parameters into the disease state mapping unit, the current medication state parameters into the medication intervention mapping unit, the current physiological regulation parameters into the physiological regulation mapping unit, and the current lifestyle disturbance parameters into the lifestyle disturbance mapping unit to obtain the disease state mapping value, medication intervention mapping value, physiological regulation mapping value, and lifestyle disturbance mapping value, respectively.

[0088] Based on the state relationships in the patient's digital twin disease progress body, the disease state mapping values, medication intervention mapping values, physiological regulation mapping values, and lifestyle disturbance mapping values ​​are spliced, weighted, and time-series labeled to generate the current disease progress state vector. Specifically, the disease state mapping values, medication intervention mapping values, physiological regulation mapping values, and lifestyle disturbance mapping values ​​are combined in a unified state arrangement order to form a continuous disease progress state sequence; the proportion of each mapping value in the continuous disease progress state sequence is standardized and adjusted according to the degree of influence of each mapping value on the changes in disease progress state; and time sequence identifiers are added to the continuous disease progress state sequence according to the time window corresponding to each mapping value to generate the current disease progress state vector.

[0089] Extract disease state samples corresponding to the historical effective control stage of patients from the patient disease course fusion dataset, perform steady-state distribution modeling on the disease state samples, and obtain the disease state distribution corresponding to the historical effective control stage of patients. The disease state samples include disease state samples, medication state samples, physiological regulation samples and lifestyle behavior disturbance samples within the historical effective control stage.

[0090] Steady-state distribution modeling includes: sorting disease state samples according to corresponding time windows to obtain historical disease state sequences; calculating the sum of absolute differences between the corresponding values ​​of disease state parameters between adjacent disease state samples in the historical disease state sequence to obtain the state change distance; calculating the difference between the maximum and minimum values ​​of each disease state parameter in the historical disease state sequence within a continuous time window to obtain the state fluctuation amplitude; calculating the length of time that disease state samples in the historical disease state sequence are continuously in the same disease stage to obtain the state duration; filtering disease state samples with state change distances below a preset distance threshold, state fluctuation amplitudes below a preset fluctuation threshold, and state durations above a preset duration threshold based on state change distance, state fluctuation amplitude, and state duration to form stable disease state regions; and generating the disease state distribution corresponding to the patient's historical effective control stages based on the parameter distribution range, parameter concentration area, and continuous stable time of the disease state samples in the stable disease state regions.

[0091] The stable state boundary is determined based on the disease course state distribution, and the current disease course state vector is mapped to the state space corresponding to the stable state boundary range to construct a stable control domain. Specifically, the stable state center is determined based on the parameter concentration region in the disease course state distribution, the stable state boundary range is determined based on the parameter distribution range in the disease course state distribution, and the stable state duration interval is determined based on the continuous stable time in the disease course state distribution; the parameter offset between each disease course state parameter in the current disease course state vector and the stable state center is calculated; the current disease course state vector is determined to be within the stable state boundary range based on the parameter offset, and the region corresponding to the current disease course state vector within the stable state boundary range is determined as the stable control domain.

[0092] In this embodiment, the construction of the virtual medication perturbation condition set specifically includes:

[0093] Extract drug names, dosages, dosing times, dosing frequencies, and concomitant drug relationships from the target patient's current medication regimen to construct normal medication conditions. Specifically, this involves generating a virtual dosing schedule consistent with the current medication regimen based on the drug names, dosages, dosing times, dosing frequencies, and concomitant drug relationships; determining the standard dosage, standard dosing sequence, and standard dosing interval for each drug within each dosing time window according to the virtual dosing schedule; and using the standard dosage, standard dosing sequence, standard dosing interval, and concomitant drug relationships as normal medication conditions.

[0094] Based on historical records of delayed medication, missed doses, make-up doses, insufficient dosage, and excessive dosage from abnormal medication use behaviors, conditions for delayed medication, missed doses, make-up doses, insufficient dosage, and excessive dosage are constructed respectively. Specifically: based on the delayed medication records in historical abnormal medication use behaviors, the time difference between the actual medication administration time and the prescribed administration time is calculated, and delayed medication conditions are constructed based on the time difference; based on the missed dose records in historical abnormal medication use behaviors, the corresponding administration time window and target drug are extracted, and conditions are constructed based on the administration time corresponding to the missed medication. The system constructs missed dose conditions based on the window and target drug; extracts the missed dose time, missed dose drug, and missed dose dosage from the missed dose records in historical abnormal medication behavior, and constructs missed dose conditions based on these information; calculates the dose gap value between the prescribed dose and the actual dose based on the dose insufficiency records in historical abnormal medication behavior, and constructs dose insufficiency conditions based on the dose gap value; calculates the excess dose difference between the actual dose and the prescribed dose based on the excess dose difference from historical abnormal medication behavior records of excessive dose, and constructs excessive dose conditions based on the excess dose difference.

[0095] Based on the dietary and exercise records in the daily life disturbance records, dietary interference conditions and exercise interference conditions are constructed respectively. Specifically, based on the dietary records in the daily life disturbance records, the time of occurrence of the meal, the duration of the meal, the amount of food consumed, and the type of food are extracted to determine the temporal overlap between the time of occurrence of the meal and the time of medication administration as prescribed. Dietary interference conditions are then constructed based on the temporal overlap, the amount of food consumed, and the type of food. Based on the exercise records in the daily life disturbance records, the time of occurrence of the exercise, the duration of the exercise, and the intensity of the exercise are extracted to determine the temporal overlap between the time of occurrence of the exercise and the time of medication administration as prescribed. Exercise interference conditions are then constructed based on the temporal overlap, the duration of exercise, and the intensity of exercise. The temporal overlap includes the time of occurrence of the meal or the time of occurrence of the exercise being within the same medication administration time window, or the duration of the meal or the duration of the exercise overlapping with the medication administration time window.

[0096] Based on the combined drug relationships in the current medication regimen and the abnormal combined drug records in historical abnormal medication behaviors, combined drug interference conditions are constructed. Specifically, the combined drug combinations, standard dosing order, and standard dosing intervals are extracted from the combined drug relationships in the current medication regimen. Records of overlapping dosing times, superimposed dosing doses, and changes in dosing order are extracted from the abnormal combined drug records in historical abnormal medication behaviors. Combined drug interference conditions are then constructed based on these records. Among these, overlapping dosing times are records where the dosing times of different drugs fall within the same dosing time window; superimposed dosing doses are records where different drug doses act together within the same dosing time window; and changes in dosing order are records where the actual dosing order deviates from the standard dosing order.

[0097] The perturbation types and deduction order of normal medication use conditions, delayed medication use conditions, missed dose conditions, make-up dose conditions, insufficient dosage conditions, excessive dosage conditions, dietary interference conditions, exercise interference conditions, and combined medication interference conditions are labeled to construct a set of virtual medication perturbation conditions. Specifically, medication behavior conditions are determined based on normal medication use conditions, delayed medication use conditions, missed dose conditions, make-up dose conditions, insufficient dosage conditions, and excessive dosage conditions; lifestyle behavior interference conditions are determined based on dietary interference conditions and exercise interference conditions; and combined medication interference conditions are determined based on combined medication interference conditions. The loading order of each type of condition in the disease progression deduction is determined based on the corresponding drug administration time window, lifestyle behavior occurrence time window, and combined medication occurrence time window. The virtual medication perturbation condition set is constructed by grouping the medication behavior conditions, lifestyle behavior interference conditions, combined medication interference conditions, and their loading order.

[0098] In this embodiment, the generation of multiple future disease progression paths specifically includes:

[0099] Extract each virtual medication perturbation condition and its loading order from the set of virtual medication perturbation conditions. Take the current disease state vector as the common initial state and establish corresponding parallel deduction channels according to each virtual medication perturbation condition. The parallel deduction channels include independent deduction channels formed by taking the current disease state vector as the common initial state, taking a single virtual medication perturbation condition as the input condition, and taking the patient's digital twin disease process as the deduction carrier.

[0100] According to the loading order, each virtual medication disturbance condition is loaded into the medication intervention mapping unit and the life behavior disturbance mapping unit in the patient's digital twin disease process;

[0101] Within the same simulation time window, the disease progression simulation unit performs disease progression simulation on each parallel simulation channel, and updates the disease state mapping value, medication intervention mapping value, physiological regulation mapping value and lifestyle disturbance mapping value in sequence according to the simulation time nodes.

[0102] The disease progression simulation includes: using the current disease state vector as the initial disease state input, reading the virtual medication perturbation conditions in the corresponding parallel simulation channel at each simulation time node; adjusting the dosing time, dosage, dosing sequence, and dosing interval in the medication intervention mapping unit based on the dosing time window, insufficient dosage condition, excessive dosage condition, combined medication interference condition, dietary interference condition, and exercise interference condition in the virtual medication perturbation conditions, and updating the medication intervention mapping value at the corresponding simulation time node; determining the dietary intake change status based on the difference in dietary intake and changes in intake time between consecutive simulation time nodes in the dietary interference condition, and determining the exercise intensity change status based on the difference in exercise intensity and changes in exercise duration between consecutive simulation time nodes in the exercise interference condition. The simulation unit adjusts the dietary fluctuation state and exercise fluctuation state in the lifestyle behavior disturbance mapping unit based on changes in dietary intake and exercise intensity, updating the lifestyle behavior disturbance mapping values ​​for the corresponding simulation time nodes. The updated medication intervention mapping values ​​and lifestyle behavior disturbance mapping values ​​are then input into the disease progression simulation unit. Combined with the disease state mapping values ​​and physiological regulation mapping values ​​from the previous simulation time node, the simulation unit calculates the changes in disease indicators, physiological indicators, and disease progression stages for the current simulation time node, updating the disease state mapping values ​​and physiological regulation mapping values ​​for the corresponding simulation time node. This update process is repeated sequentially according to the simulation time nodes to obtain the disease state mapping value, medication intervention mapping value, physiological regulation mapping value, and lifestyle behavior disturbance mapping value for each simulation time node.

[0103] For each parallel simulation channel, the disease state mapping value, medication intervention mapping value, physiological regulation mapping value, and lifestyle disturbance mapping value corresponding to each simulation time node are associated and arranged in chronological order to generate the disease state evolution sequence under the corresponding virtual medication disturbance conditions.

[0104] Each disease progression sequence is labeled and aggregated according to the corresponding virtual medication perturbation conditions to generate multiple future disease progression paths. Specifically, path type identifiers and perturbation source identifiers are added according to the virtual medication perturbation conditions corresponding to each disease progression sequence; the disease state mapping values, medication intervention mapping values, physiological regulation mapping values, and lifestyle behavior perturbation mapping values ​​are continuously arranged according to the projection time nodes, and each disease progression sequence is aggregated according to the path type identifier, perturbation source identifier, and disease progression trajectory to generate multiple future disease progression paths.

[0105] In this embodiment, the determination of the virtual disease course bifurcation point specifically includes:

[0106] Multiple future disease progression paths are time-aligned according to the same projection time node. The disease state mapping value, physiological regulation mapping value, and disease stage change state corresponding to each future disease progression path at each projection time node are extracted to form a path node state set.

[0107] A path separation analysis is performed on the set of path node states. The differences in disease state mapping values, physiological regulation mapping values, and disease stage level are calculated and weighted to obtain the path dispersion among different future disease evolution paths at the same projection time point. Specifically, the difference in disease state mapping values ​​is obtained by calculating the absolute difference between the disease state mapping values ​​of any two future disease evolution paths at the same projection time point; the difference in physiological regulation mapping values ​​is obtained by calculating the absolute difference between the physiological regulation mapping values ​​of any two future disease evolution paths at the same projection time point; and the difference in disease stage level is obtained by comparing whether the disease stages corresponding to any two future disease evolution paths at the same projection time point are consistent. When the disease stages corresponding to any two future disease evolution paths at the same projection time point are consistent, the difference in disease stage level is zero; when the disease stages are inconsistent, the difference in disease stage level is determined according to the stage level difference corresponding to the stable control stage, the fluctuation control stage, and the risk deterioration stage.

[0108] A stable control domain deviation analysis is performed on the set of path node states. The path node states corresponding to each projection time node are mapped to the stable control domain. The boundary distance, out-of-bounds time length, and out-of-bounds state intensity between the path node states and the stable control domain boundary are determined and weighted to obtain the stable domain deviation. The boundary distance is determined as follows: when the path node state is within the stable control domain boundary, the boundary distance is set to zero; when the path node state exceeds the stable control domain boundary, the difference between the state parameters exceeding the boundary in the path node state and the corresponding parameters at the stable control domain boundary is calculated, and the sum of the absolute values ​​of the differences exceeding the boundary is taken as the boundary distance. The out-of-bounds time length is obtained by counting the number of projection time nodes in which the path node state continuously exceeds the stable control domain boundary; the out-of-bounds state intensity is obtained by counting the number of state parameters exceeding the stable control domain boundary in the path node state and the corresponding out-of-bounds difference.

[0109] Risk branch generation analysis is performed on the set of path node states to identify risk paths transitioning from a stable control phase or a fluctuation control phase to a risk deterioration phase. The risk branch generation intensity is obtained by weighting and summarizing the number of risk paths, their duration, and the magnitude of risk level transitions. Specifically, the number of risk paths is obtained by counting the number of future disease progression paths transitioning from a stable control phase or a fluctuation control phase to a risk deterioration phase; the duration of risk paths is obtained by counting the number of projected time nodes in which the risk path is continuously in the risk deterioration phase; and the magnitude of risk level transitions is obtained by comparing the difference in disease stage levels before and after the risk path transitions to the risk deterioration phase.

[0110] The path dispersion, stability domain deviation, and risk branch generation intensity are used as bifurcation determination parameters. These parameters are continuously determined according to the chronological order of the simulation time nodes. The simulation time node that first consecutively meets the corresponding preset determination conditions is identified as the virtual disease bifurcation point. Specifically: the path dispersion, stability domain deviation, and risk branch generation intensity corresponding to each simulation time node are read in chronological order; it is determined whether the path dispersion, stability domain deviation, and risk branch generation intensity corresponding to at least a preset number of consecutive simulation time nodes meet the preset dispersion determination condition, the preset deviation determination condition, and the preset risk branch determination condition; the simulation time node that first consecutively meets the preset dispersion determination condition, the preset deviation determination condition, and the preset risk branch determination condition is identified as the virtual disease bifurcation point.

[0111] In this embodiment, the generation of pre-treatment monitoring results for chronic disease medication specifically includes:

[0112] Extract the projection time node, bifurcation path, and virtual medication disturbance conditions corresponding to the bifurcation point of the virtual disease course, and determine the bifurcation occurrence time window based on the projection time node. Specifically, extract the projection time node that first continuously meets the corresponding preset judgment conditions from the virtual disease course bifurcation point, determine the future disease course evolution path corresponding to the path separation, stable control domain deviation, or risk branch generation of the virtual disease course bifurcation point as the bifurcation path, determine the virtual medication disturbance conditions corresponding to the bifurcation path as the virtual medication disturbance conditions corresponding to the virtual disease course bifurcation point, and determine the bifurcation occurrence time window based on the time position of the projection time node in the disease course evolution projection.

[0113] The disease state mapping values, physiological regulation mapping values, and disease progression stage change states corresponding to consecutive projection time nodes before the bifurcation time window are extracted from the bifurcation path to generate pre-bifurcation disease state change characteristics. Specifically, the disease state mapping values, physiological regulation mapping values, and disease progression stage change states are sorted chronologically according to consecutive projection time nodes before the bifurcation time window; disease state change characteristics are determined based on the differences in disease state mapping values ​​between adjacent projection time nodes; physiological regulation change characteristics are determined based on the differences in physiological regulation mapping values ​​between adjacent projection time nodes; disease progression stage change characteristics are determined based on the stage transition order of disease progression stage change states within consecutive projection time nodes; and the disease state change characteristics, physiological regulation change characteristics, and disease progression stage change characteristics are combined to generate pre-bifurcation disease state change characteristics.

[0114] The characteristics of disease course status changes before bifurcation are correlated with the virtual medication perturbation conditions corresponding to the virtual disease course bifurcation points to determine the changes in path dispersion, stability domain deviation, and risk branch generation intensity corresponding to each virtual medication perturbation condition. Among them, the changes in path dispersion, stability domain deviation, and risk branch generation intensity are obtained by comparing the differences in path dispersion, stability domain deviation, and risk branch generation intensity between the continuous extrapolation time nodes before the bifurcation time window and the continuous extrapolation time nodes within the bifurcation time window, respectively.

[0115] Based on the changes in path dispersion, stability domain deviation, and risk branch generation intensity corresponding to each virtual medication disturbance condition, virtual medication disturbance conditions that lead to bifurcation of future disease progression paths are identified, and these identified virtual medication disturbance conditions are determined as sensitive medication factors. Specifically, the changes in path dispersion, stability domain deviation, and risk branch generation intensity corresponding to each virtual medication disturbance condition are statistically analyzed; the changes in path dispersion, stability domain deviation, and risk branch generation intensity corresponding to each virtual medication disturbance condition are numerically sorted to obtain the corresponding ranking of path dispersion, stability domain deviation, and risk branch generation intensity, and these rankings are then comprehensively compared. The virtual medication disturbance condition with the highest comprehensive ranking is extracted as the virtual medication disturbance condition that leads to bifurcation of future disease progression paths, and the corresponding changes in dosing time, dosing dosage, combined medication, dietary interference, or exercise interference are determined as sensitive medication factors.

[0116] The target patient identifier is extracted from the patient disease course fusion dataset, the chronic disease type is extracted from the disease status mapping results, the risk branch type is determined according to the disease stage change status corresponding to the virtual disease course bifurcation point, the medication monitoring prompt is generated according to the sensitive medication factors, the bifurcation occurrence time window and the risk branch type, and the target patient identifier, chronic disease type, sensitive medication factors, bifurcation occurrence time window, risk branch type and medication monitoring prompt are combined to generate the pre-monitoring results of chronic disease medication.

[0117] The determination of risk branch types includes: extracting the disease stage change status in the future disease evolution path corresponding to the virtual disease course bifurcation point; determining the target disease stage transitioning from the stable control stage or the fluctuating control stage in the disease stage change status; determining the corresponding abnormal direction of disease indicators and abnormal direction of physiological indicators based on the change direction of the disease state mapping value and the change direction of the physiological regulation mapping value corresponding to the target disease stage; when the abnormal direction of disease indicators and abnormal direction of physiological indicators correspond to persistently elevated blood glucose, persistently elevated blood pressure, persistently abnormal physiological indicators, or persistently deteriorating disease stage, the corresponding target disease stage is determined as the risk branch type.

[0118] The generation of medication monitoring prompts includes: identifying corresponding medication anomalies based on changes in dosing time, dosage, combined medication, dietary interference, or exercise interference related to sensitive medication factors. Specifically, when the sensitive medication factor is a change in dosing time, delayed or missed doses of the target drug are identified as medication anomalies; when the sensitive medication factor is a change in dosage, insufficient or excessive doses of the target drug are identified as medication anomalies; when the sensitive medication factor is a change in combined medication, the combination of drugs is identified as medication anomalies; when the sensitive medication factor is a dietary or exercise interference, the corresponding dietary or exercise records are identified as medication anomalies; determining the risk occurrence time range based on the bifurcation occurrence time window; determining the corresponding disease progression direction based on the risk branch type; generating corresponding medication monitoring prompts based on the medication anomaly, risk occurrence time range, and disease progression direction; and combining the corresponding medication monitoring prompts to generate a medication monitoring prompt.

[0119] A chronic disease medication monitoring system based on digital twins includes:

[0120] The data fusion module is used to acquire and preprocess the patient's medical history data.

[0121] The twin building module is used to construct digital twin disease profiles of patients based on a fusion dataset of patient disease histories.

[0122] The control domain module is used to generate the current disease state vector and construct a stable control domain.

[0123] The perturbation construction module is used to construct a set of virtual medication perturbation conditions;

[0124] The parallel simulation module is used to load various virtual medication perturbation conditions into the patient's digital twin disease process, perform parallel disease evolution simulation, and generate multiple future disease evolution paths;

[0125] The bifurcation identification module is used to obtain bifurcation determination parameters and determine virtual disease bifurcation points corresponding to multiple future disease evolution paths;

[0126] The results generation module is used to identify sensitive medication factors and generate pre-monitoring results for chronic disease medication.

[0127] Example 1: To verify the feasibility of this invention in practice, it was applied to a remote health management scenario for chronic diseases in a tertiary hospital. The target patients were those with long-term hypertension and diabetes undergoing home monitoring. The hospital continuously collected patients' blood pressure, blood glucose, heart rate, historical medication records, dietary records, and exercise records through wearable monitoring devices, electronic medical record systems, and mobile health terminals. This data was then combined with the current medication regimen prescribed by doctors to construct a fusion dataset of the patient's disease progression. Because some patients experience delayed or missed medication doses, or irregular eating habits, traditional monitoring methods can only provide alerts after significant abnormalities in the patient's physiological indicators. They cannot identify disease progression trends in advance, easily leading to repeated fluctuations in the patient's disease control status.

[0128] In application, the system constructs a digital twin disease history of the patient based on the patient's disease history fusion dataset and dynamically maps the patient's current disease status to generate a corresponding current disease status vector. The system constructs a stable control domain by combining the disease status distribution corresponding to the patient's historical stable control phases. Simultaneously, it constructs multiple virtual medication perturbation conditions based on the patient's recent delayed medication behavior, insufficient dosage behavior, and dietary interference behavior. Each virtual medication perturbation condition is loaded into the patient's digital twin disease history to perform parallel disease evolution simulation, thereby generating multiple future disease evolution paths. The system further performs path separation analysis, stable control domain deviation analysis, and risk branch generation analysis on these multiple future disease evolution paths to identify virtual disease bifurcation points where the disease transitions from a stable control phase to a risk deterioration phase, and pre-determines sensitive medication factors that trigger disease bifurcation. Medical staff provide early intervention and medication reminders to patients based on the medication monitoring prompts generated by the system.

[0129] Application results show that, during hospital follow-up periods lasting several months, this invention can identify the risk of disease progression caused by delayed medication, abnormal combined medication, and dietary interference in advance. Compared with traditional monitoring methods based on single monitoring data, the time for identifying disease progression risks is significantly earlier, the stability of disease progression control is improved, the duration of abnormal physiological indicators in patients is significantly shortened, and the ability of medical staff to predict changes in chronic disease risks is enhanced. This verifies the feasibility and effectiveness of this invention in the scenario of pre-emptive medication monitoring for chronic diseases.

[0130] Table 1. Performance Comparison of the Invention and Traditional Chronic Disease Medication Monitoring Methods

[0131] Indicator Categories Traditional methods Method of the present invention Risk early identification time (h) 4.6 6.1 Risk identification accuracy (%) 82.4 88.7 False alarm rate (%) 14.8 11.3 Missed report rate (%) 12.5 8.9 Duration of abnormal indicators (h) 9.2 7.6 Medication interference identification rate (%) 76.9 84.1 Medical response time (min) 38.5 31.2 Disease control stability rate (%) 79.6 85.3

[0132] As can be clearly seen from Table 1, the method of the present invention is superior to the traditional method in many indicators.

[0133] The method of this invention achieves a risk early identification time of 6.1 hours, compared to 4.6 hours for the traditional method, representing an improvement of 1.5 hours. This result demonstrates that the present invention can detect risk trends before the patient's disease course shows significant deterioration. This improvement is due to the fact that the present invention uses a digital twin of the patient's disease course to perform parallel disease course evolution simulations under different virtual medication perturbation conditions. It is no longer limited to current monitoring data but can combine factors such as delayed medication, abnormal dosage, combined medication interference, and lifestyle disturbances to predict future disease course changes, thus enabling earlier identification of risk bifurcation trends.

[0134] In terms of risk identification accuracy, the method of this invention achieves 88.7%, a 6.3% improvement compared to the traditional method's 82.4%. Simultaneously, the false alarm rate decreased from 14.8% to 11.3%, and the false negative rate decreased from 12.5% ​​to 8.9%. These results demonstrate that this invention not only improves risk identification capabilities but also enhances the stability of risk assessment. This is because traditional methods primarily rely on single abnormalities in blood pressure, blood sugar, or heart rate, making them susceptible to short-term fluctuations. In contrast, this invention constructs a disease state vector by combining disease state mapping values, physiological regulation mapping values, medication intervention mapping values, and lifestyle disturbance mapping values. Combined with deviation analysis using a stable control domain, this allows for the differentiation between normal fluctuations and persistent disease progression trends, thereby reducing false alarms and false negatives.

[0135] Regarding the duration of abnormal indicators, the method of this invention lasted 7.6 hours, while the traditional method lasted 9.2 hours, a reduction of 1.6 hours. In terms of healthcare response time, the method of this invention lasted 31.2 minutes, while the traditional method lasted 38.5 minutes, a reduction of 7.3 minutes. These results demonstrate that this invention can help healthcare professionals more quickly locate the source of risk and intervene earlier. This improvement is due to the fact that this invention not only outputs risk results but also further identifies sensitive medication factors and clarifies the corresponding changes in dosing time, dosage, dietary interference, and exercise interference. Therefore, healthcare professionals can directly address the source of risk, reducing manual screening time.

[0136] In terms of medication interference identification rate and disease course control stability rate, the method of this invention achieved 84.1% and 85.3%, respectively, while the traditional method achieved 76.9% and 79.6%, respectively. This result demonstrates that the present invention has stronger continuous monitoring capabilities in complex chronic disease management scenarios. This improvement is due to the fact that the present invention constructs a set of virtual medication perturbation conditions and continuously tracks disease course changes under different perturbation conditions through path separation analysis and risk branch generation analysis among multiple future disease course evolution paths. Therefore, it can more accurately identify potential risks in long-term disease control and improve the stability of patient disease course control.

[0137] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for monitoring medication use in chronic diseases based on digital twins, characterized in that, Includes the following steps: Acquire the patient's medical history data, preprocess it, and generate a fusion dataset of the patient's medical history. Constructing digital twin disease history data of patients based on a fusion dataset of patient disease history data; Extract the current disease status parameters from the patient's disease course fusion dataset, map the current disease status parameters to the patient's digital twin disease course to generate the current disease status vector, and construct a stable control domain by combining the disease status distribution corresponding to the patient's historical effective control stages. Based on the target patient’s current medication regimen, historical abnormal medication behavior, and records of lifestyle disturbances, a set of virtual medication disturbance conditions is constructed. Under the same current disease state vector, each virtual medication perturbation condition is loaded into the patient's digital twin disease process, parallel disease evolution inference is performed, and multiple future disease evolution paths are generated; Path separation analysis, stability control domain deviation analysis, and risk branch generation analysis are performed on multiple future disease evolution paths to obtain bifurcation determination parameters, and virtual disease bifurcation points corresponding to multiple future disease evolution paths are determined based on the bifurcation determination parameters. The determination of the virtual disease course bifurcation point specifically includes: Multiple future disease progression paths are time-aligned according to the same projection time node. The disease state mapping value, physiological regulation mapping value, and disease stage change state corresponding to each future disease progression path at each projection time node are extracted to form a path node state set. Path separation analysis is performed on the set of path node states. The differences in disease state mapping values, physiological regulation mapping values, and disease stage levels between different future disease evolution paths at the same projection time node are calculated and weighted to obtain the path dispersion. A stable control domain deviation analysis is performed on the set of path node states. The path node states corresponding to each simulation time node are mapped to the stable control domain. The boundary distance, out-of-bounds time length, and out-of-bounds state intensity between the path node states and the stable control domain boundary are determined and weighted to obtain the stable domain deviation. Risk branch generation analysis is performed on the set of path node states to identify risk paths that transition from a stable control phase or a volatility control phase to a risk deterioration phase. The risk branch generation intensity is obtained by weighted summarization based on the number of risk paths, the duration of risk paths, and the magnitude of risk level transitions. The path dispersion, stability domain deviation, and risk branch generation intensity are used as bifurcation determination parameters. The bifurcation determination parameters are continuously determined according to the order of the simulation time nodes, and the simulation time node that first continuously meets the corresponding preset determination conditions is determined as the virtual disease course bifurcation point. Based on the virtual disease course bifurcation point, the sensitive medication factors that trigger disease course bifurcation are identified, and the pre-monitoring results of chronic disease medication are generated; The generation of the pre-monitoring results for chronic disease medication specifically includes: Extract the simulation time nodes, bifurcation paths, and virtual medication disturbance conditions corresponding to the bifurcation points in the virtual disease course, and determine the bifurcation occurrence time window based on the simulation time nodes; Extract the disease state mapping value, physiological regulation mapping value and disease stage change status corresponding to the continuous extrapolation time nodes before the bifurcation time window from the bifurcation path to generate disease state change characteristics before bifurcation. The characteristics of disease course status changes before bifurcation are correlated with the virtual medication perturbation conditions corresponding to the virtual disease course bifurcation points to determine the path dispersion changes, stability domain deviation changes, and risk branch generation intensity changes corresponding to each virtual medication perturbation condition. Based on the changes in path dispersion, stability domain deviation, and risk branch generation intensity corresponding to each virtual medication perturbation condition, the virtual medication perturbation conditions that lead to the bifurcation of future disease evolution paths are identified, and the identified virtual medication perturbation conditions are determined as sensitive medication factors. The target patient identifier is extracted from the patient disease course fusion dataset, the chronic disease type is extracted from the disease status mapping results, the risk branch type is determined according to the disease stage change status corresponding to the virtual disease course bifurcation point, the medication monitoring prompt is generated according to the sensitive medication factors, the bifurcation occurrence time window and the risk branch type, and the target patient identifier, chronic disease type, sensitive medication factors, bifurcation occurrence time window, risk branch type and medication monitoring prompt are combined to generate the pre-monitoring results of chronic disease medication.

2. The method for monitoring chronic disease medication based on digital twins according to claim 1, characterized in that, The disease progression data includes basic information on chronic diseases, historical diagnosis and treatment information, historical medication information, continuous physiological monitoring information, lifestyle behavior information, and disease assessment information. The data preprocessing includes time alignment, missing data completion, indicator normalization, outlier data removal, and multi-source association mapping.

3. The method for monitoring chronic disease medication based on digital twins according to claim 1, characterized in that, The construction of the patient's digital twin pathology specifically includes: Based on the basic information on chronic diseases and historical diagnosis and treatment information in the patient disease course fusion dataset, a disease status mapping unit is established; Based on the historical medication information and current medication regimen in the patient's disease course fusion dataset, a medication intervention mapping unit is established; A physiological regulation mapping unit is established based on continuous physiological monitoring information in the patient's disease course fusion dataset; Based on the lifestyle behavior information in the patient's disease course fusion dataset, a lifestyle behavior disturbance mapping unit is established. Based on the mapping relationships between the disease state mapping unit, the medication intervention mapping unit, the physiological regulation mapping unit, and the lifestyle behavior disturbance mapping unit, a disease course evolution simulation unit is established. By associating and encapsulating the disease state mapping unit, medication intervention mapping unit, physiological regulation mapping unit, lifestyle behavior disturbance mapping unit, and disease course evolution simulation unit, a digital twin disease course of the patient is constructed.

4. The method for monitoring chronic disease medication based on digital twins according to claim 1, characterized in that, The construction of the stable control domain specifically includes: Extract the disease course data corresponding to the current time window of the target patient from the patient disease course fusion dataset, and extract the current disease status parameters, current medication status parameters, current physiological regulation parameters and current lifestyle behavior disturbance parameters based on the disease course data corresponding to the current time window to form the current disease course status parameters; Input the current disease state parameters into the disease state mapping unit, the current medication state parameters into the medication intervention mapping unit, the current physiological regulation parameters into the physiological regulation mapping unit, and the current lifestyle disturbance parameters into the lifestyle disturbance mapping unit to obtain the disease state mapping value, medication intervention mapping value, physiological regulation mapping value, and lifestyle disturbance mapping value, respectively. Based on the state associations in the patient's digital twin disease course, the disease state mapping value, medication intervention mapping value, physiological regulation mapping value, and lifestyle behavior disturbance mapping value are spliced, weighted, and time-series labeled to generate the current disease course state vector. Extract disease status samples corresponding to the historical effective control stages of patients from the patient disease course fusion dataset, perform steady-state distribution modeling on the disease status samples, and obtain the disease status distribution corresponding to the historical effective control stages of patients. The stable state boundary is determined based on the distribution of the disease course state, and the current disease course state vector is mapped to the state space corresponding to the range of the stable state boundary to construct a stable control domain.

5. A method for monitoring chronic disease medication based on digital twins according to claim 1, characterized in that, The construction of the virtual medication perturbation condition set specifically includes: Extract drug names, dosages, administration times, administration frequencies, and combination drug relationships from the target patient's current medication regimen to construct normal medication conditions; Based on the records of delayed medication, missed medication, make-up medication, insufficient dosage, and excessive dosage in historical abnormal medication behavior, conditions for delayed medication, missed medication, make-up medication, insufficient dosage, and excessive dosage are constructed respectively. Based on dietary and exercise records in the life behavior disturbance records, dietary disturbance conditions and exercise disturbance conditions were constructed respectively. Based on the combination drug relationships in the current medication regimen and the abnormal combination drug records in historical abnormal medication behaviors, construct the combination drug interference conditions; By labeling the perturbation type and the deduction order of normal medication conditions, delayed medication conditions, missed medication conditions, make-up medication conditions, insufficient dosage conditions, excessive dosage conditions, dietary interference conditions, exercise interference conditions, and combined medication interference conditions, a set of virtual medication perturbation conditions is constructed.

6. A method for monitoring chronic disease medication based on digital twins according to claim 1, characterized in that, The generation of the multiple future disease progression paths specifically includes: Extract each virtual medication perturbation condition and its loading order from the set of virtual medication perturbation conditions, take the current disease state vector as the common initial state, and establish corresponding parallel deduction channels according to each virtual medication perturbation condition. According to the loading order, each virtual medication disturbance condition is loaded into the medication intervention mapping unit and the life behavior disturbance mapping unit in the patient's digital twin disease process; Within the same simulation time window, the disease progression simulation unit performs disease progression simulation on each parallel simulation channel, and updates the disease state mapping value, medication intervention mapping value, physiological regulation mapping value and lifestyle disturbance mapping value in sequence according to the simulation time nodes. For each parallel simulation channel, the disease state mapping value, medication intervention mapping value, physiological regulation mapping value, and lifestyle disturbance mapping value corresponding to each simulation time node are associated and arranged in chronological order to generate the disease state evolution sequence under the corresponding virtual medication disturbance conditions. Each disease progression sequence is labeled and aggregated according to the corresponding virtual medication perturbation conditions to generate multiple future disease progression paths.

7. A chronic disease medication monitoring system based on digital twins, comprising the chronic disease medication monitoring method based on digital twins as described in any one of claims 1 to 6, characterized in that, include: The data fusion module is used to acquire and preprocess the patient's medical history data. The twin building module is used to construct digital twin disease profiles of patients based on a fusion dataset of patient disease histories. The control domain module is used to generate the current disease state vector and construct a stable control domain. The perturbation construction module is used to construct a set of virtual medication perturbation conditions; The parallel simulation module is used to load various virtual medication perturbation conditions into the patient's digital twin disease process, perform parallel disease evolution simulation, and generate multiple future disease evolution paths; The bifurcation identification module is used to obtain bifurcation determination parameters and determine virtual disease bifurcation points corresponding to multiple future disease evolution paths; The results generation module is used to identify sensitive medication factors and generate pre-monitoring results for chronic disease medication.

Images