General chronic disease patient dynamic monitoring and intervention management system

Abstract

According to the general chronic disease patient dynamic monitoring and intervention management system, early recognition, early intervention and continuous management of chronic disease risks are achieved through fusion of multi-source physiological signals and behavior events, self-adaptive modeling of individual baselines, segmented judgment of risk evolution trajectories, self-adaptive generation of intervention strategies and closed-loop optimization of execution feedback. The system takes a whole-process closed loop of'monitoring-evaluation-intervention-feedback-optimization 'as a core, emphasizes individualization, time sequence and interpretability, and avoids dependence of static rules and a single model, so that the management accuracy and timeliness are remarkably improved.

Description

Technical Field

[0001] This invention relates to the fields of medical and health informatics and digital health technology, and in particular to a dynamic monitoring and intervention management system for general practitioners of chronic diseases, used to achieve individualized, continuous and closed-loop health management. Background Technology

[0002] Chronic disease management has long been constrained by structural challenges such as the difficulty in effectively integrating multi-source health data, the inability to stably characterize individualized baselines, the lack of temporal discernment capabilities for risk evolution, the difficulty in adaptively optimizing intervention strategies, and the difficulty in implementing a closed-loop process. These challenges are particularly pronounced in general practice settings, where patients often have long disease durations, multiple comorbidities, and frequent daily fluctuations. Traditional models relying primarily on single-point measurements and static thresholds are unable to promptly capture individual deviations or characterize risk migration trajectories across different time scales, leading to delayed interventions, low adherence, and inefficient cross-role collaboration. Many existing solutions focus on a single data source or model, emphasizing post-event alerts while neglecting the dynamic coupling of pre-event prediction and in-event intervention, making it difficult to establish a closed-loop strategy and continuous optimization among doctors, patients, and the community.

[0003] From a data perspective, clinical visit data, follow-up records, wearable device signals, behavioral events, and environmental parameters are highly heterogeneous in both time and semantics. Conventional methods are insufficient in handling cross-modal alignment, sliding window segmentation, missing data imputation, and distribution drift, resulting in distorted feature representation and limited model generalization. From a model perspective, static rules and global models struggle to adapt to long-term individual evolution patterns. Fixed thresholds lead to a trade-off between false positives and false negatives, and there is a lack of hierarchical identification of evolutionary stages and trend-based early warning. From an intervention perspective, strategy generation often relies on experience or templates, failing to establish a quantifiable utility relationship with execution feedback, making it difficult to achieve personalized trade-offs across different patients, stages, and environments. From a system perspective, there is a tension between the full lifecycle governance requirements of data collection, transmission, storage, use, and sharing and clinical usability. Security compliance and real-time performance are difficult to balance simultaneously, impacting large-scale deployment and continuous operation.

[0004] The aforementioned problems are further amplified in the practice of regional and city-level platforms. Cross-institutional data silos, information fragmentation, and low-quality management have led to chronic disease monitoring relying more on passive reporting and post-event summarization, making it difficult to support continuous, proactive, and forward-looking management for individuals. Insufficient digital support for hierarchical diagnosis and treatment and the collaboration between medical treatment and prevention has resulted in a break in the closed-loop chain of high-risk identification, screening and intervention, full-process management, and dynamic evaluation, making it difficult to form a patient-centered, full-life-cycle health profile and a multi-disease co-management framework. Therefore, there is an urgent need for a systematic approach that is individual-centered, follows temporal evolution, focuses on strategy self-adaptation, and uses closed-loop optimization as a key tool to connect the entire process from multi-source monitoring to individualized intervention and feedback learning, and achieves stable, scalable, and auditable engineering implementation under the premise of safety and compliance. Summary of the Invention

[0005] The purpose of this invention is to provide a dynamic monitoring and intervention management system for patients with chronic diseases in general practice. To achieve this purpose, the invention is implemented through the following technical solution: A dynamic monitoring and intervention management system for patients with chronic diseases in general practice, comprising: a multi-source physiological acquisition unit, used to continuously acquire multi-source physiological signals of patients in a non-invasive manner, including but not limited to heart rate, blood pressure, respiration, skin temperature, and activity status; a behavioral event acquisition unit, used to acquire behavioral data of medication events, dietary events, exercise events, and sleep events, and align them with the physiological signals on the time axis; an environmental context acquisition unit, used to acquire environmental parameters such as temperature, humidity, air pressure, light, and noise to provide a reference for external influencing factors; an individualized baseline modeling unit, used to construct an individualized health baseline that changes over time based on individual historical data, reflecting their unique physiological rhythms and state intervals; and a temporal evolution discrimination unit, used to analyze the evolution status of the current multi-source data based on the individualized baseline. The system comprises the following components: a state discrimination unit, which identifies stages such as steady state, slight deviation, significant deviation, and acute abnormality; a risk quantification and grading unit, which calculates individual risk quantification values ​​based on evolutionary states and dynamically grades them to achieve precise differentiation of risk levels; an adaptive intervention strategy generation unit, which generates individualized intervention strategies based on risk quantification values ​​and evolutionary states, covering multiple aspects such as medication, exercise, diet, sleep, environment, and follow-up visits; an execution and feedback collection unit, which distributes intervention strategies to patient-side terminals and collects execution feedback data, including completion rate, timestamps, and subjective feelings; a closed-loop optimization unit, which updates individualized baselines and strategy parameters based on execution feedback data, enabling the system to have self-learning and adaptive capabilities; and a security and compliance control unit, which performs full lifecycle security and compliance control on data collection, transmission, storage, and use to ensure privacy and data security. The system defines a complete closed-loop architecture, forming an end-to-end process from multi-source data collection to individualized modeling, risk assessment, strategy generation, execution feedback and optimization. It also incorporates security and compliance safeguards, enabling the system to achieve continuous, dynamic and traceable chronic disease management in general practice scenarios, thus solving the problems of fragmentation and lack of closed loop in traditional systems.

[0006] Furthermore, the multi-source physiological acquisition unit includes: a photoplethysmography (PPG) sensor, an electrocardiogram (ECG) sensor, a piezoresistive blood pressure sensor, an impedance respiration sensor, a skin temperature sensor, and a triaxial accelerometer; the multi-source physiological acquisition unit is aggregated at a sampling frequency. Synchronous sampling is performed, and physiological fragment sequences are generated by timestamp alignment and sliding window segmentation. ,in Indicates the first The system generates physiological feature vectors for each time window. These vectors, x_i, are composed of temporal, frequency, and morphological features, and are used to characterize comprehensive information about cardiovascular, respiratory, autonomic nervous, and activity states, thereby depicting the patient's current state in multiple dimensions. Through simultaneous sampling and feature fusion from multiple sensors, the system can comprehensively capture cardiovascular, respiratory, autonomic nervous, and activity information. Combining temporal, frequency, and morphological features improves the accuracy and robustness of state description, providing high-quality input for subsequent individualized modeling and risk assessment.

[0007] Furthermore, the individualized baseline modeling unit employs an Individual Temporal Convolutional Network (ITCN) to construct an individualized baseline function. The individualized baseline function The information is obtained as follows: An individual's historical physiological sequence X and its corresponding environmental context sequence E are input into the ITCN. The ITCN consists of L layers of dilated causal convolutions, with the kernel of the l-th layer being... Expansion rate The output is in a hidden state. ITCN integrates cross-scale time patterns through a gating mechanism to generate individual time... Baseline characterization The individualized baseline modeling unit further introduces learnable rhythmic terms. With seasonal items To obtain the individualized baseline function ,in Expanded by Fourier basis functions, Based on annual and weekly cycle basis functions, the baseline is characterized by the contribution of intraday and weekly / monthly rhythms to the baseline, making the baseline more closely reflect individual physiological patterns. The introduction of ITCN and rhythm and seasonal terms enables the baseline to not only reflect long-term trends but also capture intraday and weekly / monthly rhythm variations, significantly improving the individualization and stability of the baseline and reducing misjudgments caused by rhythm differences.

[0008] Furthermore, the temporal evolution discrimination unit determines the current physiological segment. Evolutionary state determination includes: calculating the deviation sequence. ; Calculate the trend slope sequence ,in For trend windows; calculate multi-scale volatility sequences ,in For fluctuation window; , and Concatenate into evolutionary feature vectors Attention-gated recursive units (AGRUs) are used to... Encode to obtain the hidden state Based on hidden state Calculate the probability distribution of evolutionary states ,in These are learnable parameters; the evolutionary states include four categories: steady state, slight deviation, significant deviation, and acute anomaly, each corresponding to a different stage of risk evolution, enabling a fine-grained characterization of the risk migration process. By comprehensively utilizing deviation, trend slope, and volatility, and capturing temporal dependencies through AGRU encoding, the system can hierarchically identify risk evolution stages, providing a basis for accurate classification and early warning.

[0009] Furthermore, the risk quantification and classification unit calculates the individual risk quantification value. This includes: based on the probability distribution of evolutionary states Calculate state risk Based on deviation sequence Calculate the magnitude risk Based on trend slope sequence Calculate trend risk Based on multi-scale volatility sequences Calculate volatility risk Comprehensive calculation of risk quantification value ,in , , , For learnable weights; according to Classification: When When it is low risk, At the time, it was considered a medium-risk period. At that time, it was considered a medium-to-high risk level. The time is high risk, among which , , The system uses dynamic thresholds to achieve multi-dimensional quantification and dynamic classification of risks. It integrates four types of risk measures: uncertainty, magnitude, trend, and volatility, and uses learnable weights to achieve individualized comprehensive scoring. The dynamic thresholds make the classification more closely match the distribution of individuals and groups, improving the accuracy and flexibility of early warning.

[0010] Furthermore, the dynamic threshold , , The update methods include: calculating the historical distribution statistics of individual risk quantification values. and ; Calculate the distribution statistics of the group risk quantification value and Thresholds are updated based on the relative positions of individuals and groups. ,in , , The adaptive coefficient; the dynamic threshold update period Adaptive adjustment based on risk volatility: ,in Based on the update cycle, This is a volatility sensitivity coefficient used to accelerate threshold updates when risk fluctuations are significant, ensuring the timeliness of the classification. The threshold is updated based on the relative position of individual and group statistics, and the update cycle is adaptively adjusted according to volatility. This allows the system to respond quickly during periods of high risk and remain stable during periods of stability, avoiding strategy oscillations caused by frequent threshold switching.

[0011] Furthermore, the adaptive intervention strategy generation unit generates the strategy based on the risk quantification value. With evolutionary state Generate individualized intervention strategies, including: constructing a strategy knowledge graph K, where nodes represent intervention actions and edges represent synergistic and mutually exclusive relationships between actions; based on... and Perform a restricted path search in K to generate a set of candidate strategies. ; Calculate the candidate strategy set Each strategy utility rating ,in The efficacy score is predicted based on historical feedback. To implement the feasibility score, Score the patient burden. , , Assign weights; select the strategy with the highest utility score. As a current intervention strategy, the intervention strategy includes at least one or more combinations of medication adherence improvement, exercise prescription adjustment, dietary structure adjustment, sleep optimization, environmental exposure control, and follow-up visit reminders, to achieve personalized intervention under a multi-objective trade-off. The synergistic and mutually exclusive relationships of intervention actions are expressed through a knowledge graph, and the optimal strategy is selected within a multi-objective utility scoring framework, making the intervention both efficient and feasible, significantly improving patient adherence and intervention effectiveness.

[0012] Furthermore, the execution and feedback acquisition unit includes: a strategy distribution module, used to distribute intervention strategies. Convert to a list of executable tasks for the patient. and for each task Allocate execution time windows and reminder methods; execution record module, used to collect data on patient task completion. Timestamps, completion status, and subjective feedback; a compliance calculation module for calculating task compliance rate. ,in The function is an indicator; the execution and feedback acquisition unit further calculates the intervention effect index. ,in This is a quantified value of risk before intervention. The compliance calculation module stores A_k and ΔR together for subsequent strategy optimization, forming a quantifiable feedback loop. It decomposes the strategy into traceable tasks and records execution details and subjective feedback. Combined with compliance rate and intervention effect indicators, it provides reliable causal evidence for strategy optimization and improves the system's learning and adaptation capabilities.

[0013] Furthermore, the closed-loop optimization unit updates the individualized baseline and policy parameters based on execution feedback data, including: updating the ITCN parameters and rhythm terms in the individualized baseline modeling unit. Seasonal items To reflect the new steady state after intervention; update the weights in the risk quantification and classification units. , , , With dynamic threshold , , To reflect changes in individual risk structure; update the strategy knowledge graph K and utility score weights in the adaptive intervention strategy generation unit. , , This reflects the true efficacy and burden of intervention actions. The closed-loop optimization unit employs an incremental learning strategy, updating only a subset of parameters related to feedback. This ensures stability while improving learning efficiency, enabling the system to continuously evolve. By updating only relevant parameters through incremental learning, the system maintains stable operation while continuously learning, allowing the baseline and strategy to adaptively evolve with changes in patient status and behavior.

[0014] Furthermore, the security and compliance control unit includes: a data acquisition minimization module, used to collect only the minimum dataset necessary for monitoring and intervention; a data desensitization and anonymization module, used to desensitize identity information and sensitive data, generating anonymized identifiers; a transmission encryption module, used to encrypt data transmission and verify integrity using national cryptographic algorithms SM2 / SM3 / SM4; an access control module, used for access control based on roles and the principle of least privilege; an audit trail module, used for full-link auditing of data access and operations; and a data retention and destruction strategy module, used for data retention and expiration destruction according to regulatory requirements. This security and compliance control unit spans the entire lifecycle of data acquisition, transmission, storage, use, and sharing, ensuring patient privacy and data security, and meeting clinical and regulatory requirements. This full lifecycle security and compliance control mechanism ensures that data is rigorously protected at every stage, enhancing patient trust and improving the system's deployability and compliance within medical institutions.

[0015] This invention provides a dynamic monitoring and intervention management system for general practitioners of chronic diseases, which has the following beneficial effects:

[0016] Based on the full absorption of multi-source health data, individualized modeling, and closed-loop management concepts, this system achieves precision, timeliness, and sustainable optimization of general practitioner chronic disease management through systematic innovation.

[0017] First, the multi-source physiological acquisition unit covers multi-dimensional signals such as cardiovascular, respiratory, autonomic nervous and activity signals. Combined with behavioral events and environmental context, it forms a high-fidelity physiological segment sequence with time alignment and sliding window segmentation, providing a rich, robust and interpretable feature foundation for subsequent modeling.

[0018] Secondly, the individualized baseline modeling unit introduces individual temporal convolutional networks and rhythm and seasonal terms, which can capture long-term dependencies at multiple scales and characterize the contribution of intraday and weekly / monthly rhythms to individual states, significantly improving the stability and transferability of the baseline and reducing the systematic errors of false alarms and false negatives.

[0019] Third, the temporal evolution discrimination unit integrates deviation, trend slope and multi-scale volatility, and uses attention-gated recursive coding to characterize the risk migration trajectory. It also achieves hierarchical identification from steady state to acute anomaly with four types of evolution states, providing interpretable basis for risk classification and early warning.

[0020] Fourth, the risk quantification and classification unit integrates four types of risk measures: uncertainty, magnitude, trend and fluctuation. It uses learnable weights to achieve individualized comprehensive scoring and dynamically updates the threshold based on the relative position of the individual and the group. This allows risk judgment to reflect both individual specificity and changes in group distribution, thereby improving the sensitivity and specificity of early warning.

[0021] Fifth, the dynamic threshold update mechanism adaptively adjusts the update cycle based on risk volatility, accelerating iteration during periods of high risk and reducing disturbances during periods of stability. This ensures the model responds quickly to changes while avoiding strategy oscillations caused by frequent threshold switching.

[0022] Sixth, the adaptive intervention strategy generation unit uses a strategy knowledge graph to express the synergistic and mutually exclusive relationships between actions. Through utility scoring, it selects the optimal strategy under multi-objective trade-offs, realizing the calculability and personalization of combined interventions such as improved medication adherence, adjustment of exercise prescriptions, adjustment of dietary structure, optimization of sleep, control of environmental exposure, and reminders for follow-up visits. This significantly improves the feasibility of interventions and the predictability of efficacy.

[0023] Seventh, the execution and feedback collection unit breaks down the strategy into a traceable task list, collects completion rate, timestamps and subjective feedback, calculates compliance rate and intervention effect indicators, and incorporates efficacy and burden into a unified evaluation framework to provide high-quality causal evidence for strategy optimization.

[0024] Eighth, the closed-loop optimization unit uses incremental learning to update only a subset of parameters related to feedback, which improves learning efficiency while ensuring system stability, so that the baseline and strategy continue to evolve with changes in patient status and behavior, forming a data-driven virtuous cycle.

[0025] Ninth, the security and compliance control unit runs through the entire data lifecycle, using mechanisms such as minimized collection, anonymization, encrypted transmission, access control, and audit trail to ensure privacy and data security, meet clinical deployment and regulatory compliance requirements, and enhance patient trust and institutional adoption.

[0026] Tenth, the system adopts a modular architecture to achieve high cohesion and low coupling, facilitating flexible deployment and expansion in multiple scenarios such as general practice and specialty, community and hospital, home and outpatient care. It supports collaborative management of multiple diseases, devices, and roles, promoting a paradigm shift in chronic disease management from experience-driven to data-driven, from fragmented management to full-cycle closed-loop management, and from single-disease management to multi-disease co-management. Through the organic unity of multi-source fusion, individualized baselines, time-series discrimination, dynamic grading, adaptive intervention, and closed-loop optimization, the system achieves synergistic enhancement in the foresight of risk identification, the adaptability of intervention strategies, and the sustainability of the management process, providing a comprehensive solution for continuous management of chronic diseases in general practice that is scientific, engineered, and feasible. Attached Figure Description

[0027] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0028] Figure 1 This is a flowchart illustrating the overall closed-loop operation of the system of the present invention. Figure 2 This is a flowchart of the individualized baseline modeling and evolutionary discrimination process of this invention; Figure 3 This is a flowchart illustrating the risk quantification and dynamic classification process of this invention. Figure 4 This is a flowchart illustrating the adaptive intervention strategy generation and execution feedback process of the present invention. Detailed Implementation

[0029] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses consistent with some aspects of this disclosure as detailed in the appended claims.

[0030] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0031] Usage: Using the multi-source physiological acquisition unit: During the patient's daily activities, activate the multi-source physiological acquisition unit. It synchronously samples data from the PPG sensor, ECG sensor, piezoresistive blood pressure sensor, impedance respiration sensor, skin temperature sensor, and triaxial accelerometer at a preset sampling frequency set. During acquisition, the system automatically timestamps and segments various signals using a sliding window, generating a sequence of physiological segments. Each segment contains time-domain, frequency-domain, and morphological features, used to characterize comprehensive information about cardiovascular, respiratory, autonomic nervous, and activity states.

[0032] Use of the behavioral event collection unit: Patients manually or automatically record medication events, dietary events, exercise events, and sleep events on terminal devices or medical care terminals. The system aligns these behavioral data with physiological signals on the timeline to form joint temporal data of behavior and physiology, providing behavioral dimension input for subsequent individualized modeling.

[0033] Use of the environmental context acquisition unit: The system continuously collects temperature, humidity, air pressure, light and noise parameters through built-in or external environmental sensors, and stores these environmental data and physiological and behavioral data synchronously so that the influence of the external environment can be considered in the modeling and discrimination process.

[0034] The use of the Individualized Baseline Modeling Unit (ITCN): Patient historical physiological segments and corresponding environmental context sequences are input into the Individual Temporal Convolutional Network (ITCN). After processing through multiple layers of dilated causal convolutions and gating mechanisms, the individual's temporal baseline modeling unit is generated. The system further combines learnable rhythmic and seasonal terms to construct an individualized baseline function that reflects the patient's unique physiological rhythms and long-term state intervals.

[0035] The temporal evolution discrimination unit is used to calculate the deviation sequence, trend slope sequence, and multi-scale volatility sequence for the current physiological segment, and concatenate them into an evolutionary feature vector. An attention-gated recursive unit is used to encode the feature vector, obtain the hidden state, and calculate the probability distribution of the evolutionary state, thereby determining whether the patient is in a steady state, mild deviation, significant deviation, or acute abnormality stage.

[0036] The system utilizes risk quantification and classification units: State risk is calculated based on the probability distribution of evolving states. This is combined with deviation magnitude risk, trend risk, and volatility risk, and learned weights are used for comprehensive calculation to obtain an individual risk quantification value. The system then classifies this value into low-risk, medium-risk, medium-high-risk, and high-risk levels based on dynamic thresholds, achieving precise differentiation of risk levels.

[0037] The adaptive intervention strategy generation unit searches for a set of candidate strategies that meet the criteria in the strategy knowledge graph based on risk quantification and evolution status. Utility scores are calculated for each candidate strategy, and the strategy with the highest score is selected as the current intervention strategy, taking into account efficacy, feasibility, and patient burden. The strategy may include improving medication adherence, adjusting exercise prescriptions, modifying dietary structure, optimizing sleep, controlling environmental exposure, and reminding patients to return for follow-up appointments.

[0038] The execution and feedback collection unit converts the generated intervention strategy into a list of tasks that patients can perform, and assigns execution time windows and reminder methods. After patients complete each task on the terminal, the system records the timestamp, completion rate, and subjective feedback, calculates the task compliance rate and intervention effect indicators, and stores the compliance rate and effect indicators in association for subsequent optimization.

[0039] The closed-loop optimization unit is used as follows: Based on execution feedback data, the system updates the ITCN parameters, rhythmic terms, and seasonal terms in the individualized baseline modeling unit, adjusts the weights and dynamic thresholds in the risk quantification and grading unit, and optimizes the knowledge graph and utility score weights in the strategy generation unit. The update process uses incremental learning, adjusting only a subset of parameters related to feedback to ensure system stability and learning efficiency.

[0040] Use of the security and compliance control unit: During the data acquisition phase, the system follows the principle of minimization, collecting only necessary data. In the data transmission and storage phases, identity information is anonymized and desensitized, and encryption algorithms are used to ensure transmission security and data integrity. Access control is implemented based on the principle of role-based access and least privilege, with end-to-end auditing of data access and operations. Data retention and expiration destruction are performed in accordance with regulatory requirements, ensuring patient privacy and data security throughout the entire system lifecycle.

[0041] Example:

[0042] Example 1: Quiet Home Environment

[0043] In a quiet home environment, patients wear wearable devices integrating multi-source physiological acquisition units, including a photoplethysmography (PPG) sensor, an electrocardiogram (ECG) sensor, a piezoresistive blood pressure sensor, an impedance respiration sensor, a skin temperature sensor, and a triaxial accelerometer. The devices synchronously sample according to a preset sampling frequency set. The system performs timestamp alignment and sliding window segmentation on the signals to generate a sequence of physiological segments. Temporal, frequency, and morphological features are extracted from each segment to characterize comprehensive information about cardiovascular, respiratory, autonomic nervous, and activity states. A behavioral event acquisition unit records medication, dietary, exercise, and sleep events via a patient application, aligning them with physiological signals on the timeline. An environmental context acquisition unit continuously collects indoor temperature, humidity, air pressure, light intensity, and noise, forming comprehensive time-series data on the individual's internal state and external environment. The individualized baseline modeling unit inputs historical physiological segment sequences and environmental context sequences into the individual temporal convolutional network (ITCN). After processing through multi-layer dilated causal convolutions and gating mechanisms, it generates the individual's baseline representation at time t, and superimposes learnable rhythmic and seasonal terms to construct an individualized baseline function that reflects the influence of intraday and weekly / monthly rhythms on the baseline. The temporal evolution discrimination unit calculates the deviation sequence, trend slope sequence, and multi-scale volatility sequence of the current physiological segment, concatenates them into an evolutionary feature vector, encodes it using attention-gated recursive units, obtains the hidden state, and calculates the evolutionary state probability distribution to determine whether the patient is in a steady state, mild deviation, significant deviation, or acute abnormal stage. The risk quantification and classification unit calculates the state risk based on the evolutionary state probability distribution, combines the deviation magnitude risk, trend risk, and volatility risk, and uses learnable weights to comprehensively obtain the individual risk quantification value, and classifies it into low risk, medium risk, medium-high risk, and high risk levels according to dynamic thresholds. The adaptive intervention strategy generation unit searches for candidate strategies in the strategy knowledge graph based on risk quantification values ​​and evolution status, calculates utility scores, and selects the strategy with the highest score. These strategies cover areas such as medication adherence improvement, exercise prescription adjustment, dietary structure adjustment, sleep optimization, environmental exposure control, and follow-up visit reminders. The execution and feedback collection unit transforms the strategies into a list of executable tasks for patients, assigns execution time windows and reminder methods, records completion rates, timestamps, and subjective feedback, calculates task adherence rates and intervention effect indicators, and stores them for subsequent optimization. The closed-loop optimization unit, based on execution feedback data, uses an incremental learning strategy to update the ITCN parameters, rhythmic terms, and seasonal terms in the individualized baseline modeling unit, updates the weights and dynamic thresholds in the risk quantification and grading unit, and updates the knowledge graph and utility score weights in the strategy generation unit, ensuring continuous evolution of the system during stable operation.During the data collection phase, the security and compliance control unit follows the principle of minimization, desensitizes and anonymizes identity information and sensitive data, uses encryption algorithms to ensure transmission security and data integrity, implements access control based on the principle of least privilege, conducts full-link auditing of data access and operations, and performs data retention and expiration destruction in accordance with regulatory requirements to ensure privacy and data security throughout the entire lifecycle.

[0044] Example 2: Public Activity Environment of a Community Day Care Center

[0045] In the public activity environment of a community day care center, patients wear lightweight multi-source physiological data acquisition units for continuous monitoring. The devices simultaneously collect signals such as heart rate, blood pressure, respiration, skin temperature, and activity intensity. The system generates physiological segment sequences through timestamp alignment and sliding window segmentation, extracting multidimensional features to characterize individual states within a group activity context. A behavioral event acquisition unit, assisted by caregivers, records medication events, dietary events, exercise events, and sleep events, aligning them with physiological signals to supplement group activity participation. An environmental context acquisition unit collects activity room temperature, humidity, air pressure, light, and noise, focusing on the contextual impact of crowd gathering and noise fluctuations on individual states. A personalized baseline modeling unit uses patients' historical home and community data to construct a personalized baseline function, focusing on characterizing the perturbation patterns of baselines caused by daytime activity rhythms and social interactions. A temporal evolution discrimination unit calculates the deviation, trend slope, and multi-scale volatility of the current physiological segment during activity, comprehensively encodes these parameters, and determines the evolutionary state, identifying mild and significant deviations in contexts involving overlapping social and activity activities. The risk quantification and grading unit calculates individual risk quantification values ​​based on evolutionary status and risk measurement, and grades them using dynamic thresholds to ensure accurate risk level differentiation even in group activity scenarios. The adaptive intervention strategy generation unit selects low-interference, highly executable intervention actions from the strategy knowledge graph based on risk level and evolutionary status, such as short-term relaxation training, hydration reminders, gait rhythm adjustment, and environmental noise exposure control, and distributes them as group reminders. The execution and feedback collection unit translates the strategies into micro-tasks in the activity schedule, records participation status, completion rate, and subjective feelings, calculates compliance rate and intervention effect indicators, and uses them to optimize group intervention plans. The closed-loop optimization unit updates individualized baselines and strategy parameters based on feedback data, focusing on optimizing the identification and intervention weights of activity-related fluctuations to improve the accuracy of judgment and intervention comfort in public environments. The security and compliance control unit strengthens anonymization and access control in scenarios where multiple people share devices, ensuring the isolation and auditing of individual data throughout the entire process of collection, transmission, storage, and use, meeting the management and regulatory requirements of community organizations.

[0046] Example 3: Open-plan office workspace environment

[0047] In an open-plan office environment, patients wear miniaturized multi-source physiological acquisition units. The devices continuously acquire cardiovascular and activity-related signals in a low-power mode. The system generates physiological segment sequences adapted to alternating periods of sedentary activity and intermittent activity through sliding window segmentation and feature extraction. A behavioral event acquisition unit allows patients to manually or automatically record medication, dietary, exercise, and sleep events, aligning them with their work schedule and supplementing with sedentary reminders and lunch break activities. An environmental context acquisition unit collects office temperature, humidity, air pressure, light, and noise, focusing on monitoring the potential impact of air conditioning airflow, noise interference, and changes in light on the patient's condition. A personalized baseline modeling unit integrates historical data from both home and office to construct a baseline function, highlighting the shaping effect of workday rhythms and task load on the baseline. A temporal evolution discrimination unit calculates deviation, trend slope, and multi-scale volatility under the background of prolonged sitting and short-term activity switching, identifying mild deviations caused by prolonged sitting, attention load, and emotional fluctuations. A risk quantification and grading unit calculates risk quantification values ​​by comprehensively considering state uncertainty, amplitude, trend, and volatility measures, and grades them according to dynamic thresholds to ensure timely and non-intrusive warnings in the work environment. The adaptive intervention strategy generation unit prioritizes low-intrusion interventions, such as standing reminders, eye and neck / shoulder relaxation exercises, segmented walking, and hydration suggestions, and integrates with calendar tools for execution. The execution and feedback collection unit embeds the strategies into the office schedule, records task triggering and completion trajectories, collects subjective feedback, and calculates compliance rates and intervention effectiveness indicators, providing a basis for optimizing the strategy's adaptability in the office environment. The closed-loop optimization unit updates the parameters of the individualized baseline modeling unit and the strategy generation unit based on feedback data, enhancing the ability to identify sedentary and task load patterns, and improving the personalization and acceptability of interventions. In scenarios involving shared office equipment and network access, the security and compliance control unit strengthens data minimization, anonymization, and encrypted transmission, implementing fine-grained access control and operational auditing to ensure employee privacy and organizational compliance.

[0048] Example 4: Hospital outpatient waiting environment

[0049] In the hospital outpatient waiting environment, patients wear multi-source physiological acquisition units for short-range continuous monitoring. The devices simultaneously collect cardiovascular, respiratory, and activity signals. The system generates high-temporal-resolution physiological segment sequences through timestamp alignment and sliding window segmentation to reflect the impact of emotional stress and postural changes on physiological state during the waiting period. Behavioral event acquisition units record medication events, dietary events, exercise events, and sleep events, supplementing with behavioral nodes such as registration, waiting, and follow-up appointment arrangements. Environmental context acquisition units collect temperature, humidity, air pressure, light, and noise in the waiting area, focusing on the contextual impact of crowd movement, call-number announcements, and device prompts on individual states. Personalized baseline modeling units utilize patients' historical home, community, and office data to construct baseline functions, focusing on characterizing physiological and emotional baseline shifts in the medical context. Temporal evolution discrimination units calculate deviation, trend slope, and multi-scale volatility during the waiting process, identifying mild and significant deviations caused by stress, static posture, and noise stimulation. The risk quantification and grading unit calculates individual risk quantification values ​​using multi-source risk metrics and grades them according to dynamic thresholds, providing priority references for the medical team. The adaptive intervention strategy generation unit selects immediately executable non-pharmacological interventions from the strategy knowledge graph based on risk level and evolution status, such as rhythmic breathing guidance, relaxation prompts, hydration, and toilet reminders, and links them with the queuing system to avoid missed appointments. The execution and feedback collection unit transforms strategies into micro-tasks during the waiting period, records triggering and completion status, collects subjective feedback, and calculates compliance rates and intervention effectiveness indicators, providing a basis for subsequent outpatient follow-up and intervention optimization. The closed-loop optimization unit updates individualized baselines and strategy parameters based on feedback data, strengthens the recognition of emotion and postural patterns in the medical context, and improves the contextual sensitivity of warnings and the timeliness of interventions. The security and compliance control unit strictly implements minimal data collection, anonymization, encrypted transmission, access control, and end-to-end auditing within the medical institution environment to ensure patient privacy and medical data security, meeting clinical deployment and regulatory requirements.

[0050] Example 5: Business Travel and Long-Distance Flight Environment

[0051] In long-haul flights across time zones, patients wear portable multi-source physiological data acquisition units. These devices continuously acquire cardiovascular, respiratory, skin temperature, and activity signals in an anti-interference mode. The system generates physiological segment sequences adapted to the confined cabin environment and sedentary lifestyle through sliding window segmentation and feature extraction. A behavioral event acquisition unit records medication, dietary, exercise, and sleep events, supplementing with behavioral nodes such as boarding, meals, takeoff, landing, and transfers. An environmental context acquisition unit collects cabin temperature, humidity, air pressure, light, and noise, focusing on the combined effects of low humidity, air pressure changes, cabin noise, and nighttime lighting on the patient's condition. A personalized baseline modeling unit uses the patient's historical home, community, office, and waiting room data to construct a baseline function, introducing rhythm and seasonal terms to characterize the perturbation of diurnal rhythms across time zones. A temporal evolution discrimination unit calculates deviation, trend slope, and multi-scale volatility at different stages of flight, identifying mild and significant deviations caused by sedentary lifestyles, disrupted sleep rhythms, and mild hypoxia. The risk quantification and grading unit calculates individual risk quantification values ​​based on a comprehensive measure of uncertainty, magnitude, trend, and volatility, and grades these values ​​according to dynamic thresholds, providing a basis for non-pharmacological interventions during the flight. The adaptive intervention strategy generation unit selects appropriate intervention actions from a strategy knowledge graph based on risk level and evolutionary status, such as segmented standing and ankle pump exercises, hydration and low-salt diet recommendations, sleep rhythm guidance, and light exposure control, providing gentle reminders via vibration and voice. The execution and feedback collection unit transforms the strategies into a flight task list, records completion rates and subjective feedback, and calculates compliance rates and intervention effectiveness indicators, providing data support for follow-up and re-baselineing upon arrival at the destination. The closed-loop optimization unit updates individualized baselines and strategy parameters based on feedback data, enhancing the joint identification capabilities of barometric pressure changes, sedentary behavior, and cross-time zone rhythm shifts, improving management continuity in special travel scenarios. The safety and compliance control unit strengthens data minimization, anonymization, and encrypted transmission under cross-border and in-flight network conditions, implementing strict access control and operational auditing to ensure privacy and data security in business travel scenarios.

[0052] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A dynamic monitoring and intervention management system for general practitioners of chronic diseases, characterized in that: include: A multi-source physiological acquisition unit is used to continuously acquire multi-source physiological signals from patients in a non-invasive manner; The behavioral event collection unit is used to collect behavioral data on medication events, dietary events, exercise events, and sleep events. The environmental context acquisition unit is used to collect environmental parameters such as temperature, humidity, air pressure, light intensity, and noise. The individualized baseline modeling unit is used to construct individualized health baselines based on individual historical data; The temporal evolution discrimination unit is used to discriminate the evolutionary state of the current multi-source data based on an individualized baseline; the risk quantification and classification unit is used to calculate the individual risk quantification value and classify it according to the evolutionary state. An adaptive intervention strategy generation unit is used to generate individualized intervention strategies based on risk quantification values ​​and evolution status. The execution and feedback acquisition unit is used to distribute intervention strategies to patient-side terminals and collect execution feedback data; the closed-loop optimization unit is used to update individualized baselines and strategy parameters based on execution feedback data; and the security and compliance control unit is used to perform full lifecycle security and compliance control over data acquisition, transmission, storage, and use.

2. The dynamic monitoring and intervention management system for general practitioners with chronic diseases according to claim 1, characterized in that, The multi-source physiological acquisition unit includes: a photoplethysmography (PPG) sensor, an electrocardiogram (ECG) sensor, a piezoresistive blood pressure sensor, an impedance respiration sensor, a skin temperature sensor, and a triaxial accelerometer; the multi-source physiological acquisition unit is aggregated at a sampling frequency. Synchronous sampling is performed, and physiological fragment sequences are generated by timestamp alignment and sliding window segmentation. ,in Indicates the first Physiological feature vectors for each time window; the physiological feature vectors It consists of time-domain features, frequency-domain features, and morphological features, and is used to characterize comprehensive information about cardiovascular, respiratory, autonomic nervous, and activity states.

3. The dynamic monitoring and intervention management system for general practitioners with chronic diseases according to claim 1, characterized in that, The individualized baseline modeling unit uses the Individual Temporal Convolutional Network (ITCN) to construct the individualized baseline function. The individualized baseline function The information is obtained as follows: An individual's historical physiological sequence X and its corresponding environmental context sequence E are input into the ITCN. The ITCN consists of L layers of dilated causal convolutions, with the kernel of the l-th layer being... Expansion rate The output is in a hidden state. ITCN integrates cross-scale time patterns through a gating mechanism to generate individual time... Baseline characterization The individualized baseline modeling unit further introduces learnable rhythmic terms. With seasonal items To obtain the individualized baseline function ,in Expanded by Fourier basis functions, It is derived from annual and weekly cycle basis functions and is used to characterize the contribution of intraday and weekly / monthly rhythms to the baseline.

4. The dynamic monitoring and intervention management system for general practitioners with chronic diseases according to claim 1, characterized in that, The temporal evolution discrimination unit analyzes the current physiological segment. Evolutionary state determination includes: calculating the deviation sequence. ; Calculate the trend slope sequence ,in For trend windows; calculate multi-scale volatility sequences ,in For fluctuation window; , and Concatenate into evolutionary feature vectors Attention-gated recursive units (AGRUs) are used to... Encode to obtain the hidden state Based on hidden state Calculate the probability distribution of evolutionary states ,in These are learnable parameters; the evolutionary states include four categories: steady state, slight deviation, significant deviation, and acute abnormality, each corresponding to a different stage of risk evolution.

5. The dynamic monitoring and intervention management system for general practitioners with chronic diseases according to claim 1, characterized in that, The risk quantification and classification unit calculates the individual risk quantification value. Including: based on evolutionary state probability distribution Calculate state risk ,in A set of evolutionary states; based on the deviation sequence Calculate the magnitude risk Based on trend slope sequence Calculate trend risk Based on multi-scale volatility sequences Calculate volatility risk Comprehensive calculation of risk quantification value ,in , , , For learnable weights; according to Classification: When When it is low risk, At the time, it was considered a medium-risk period. At that time, it was considered a medium-to-high risk level. It is a high-risk time, among which , , This is a dynamic threshold.

6. The dynamic monitoring and intervention management system for general practitioners with chronic diseases according to claim 5, characterized in that, The dynamic threshold , , The update methods include: calculating the historical distribution statistics of individual risk quantification values. and ; Calculate the distribution statistics of the group risk quantification value and Thresholds are updated based on the relative positions of individuals and groups. ,in , , The adaptive coefficient; the dynamic threshold update period Adaptive adjustment based on risk volatility: ,in Based on the update cycle, This is a volatility sensitivity coefficient, used to accelerate threshold updates when risk fluctuations are significant.

7. The dynamic monitoring and intervention management system for general practitioners with chronic diseases according to claim 1, characterized in that, The adaptive intervention strategy generation unit generates the strategy based on the risk quantification value. With evolutionary state Generate individualized intervention strategies, including: constructing a strategy knowledge graph K, where nodes represent intervention actions and edges represent synergistic and mutually exclusive relationships between actions; based on... and Perform a restricted path search in K to generate a set of candidate strategies. ; Calculate the candidate strategy set Each strategy utility rating The efficacy score is predicted based on historical feedback. To implement the feasibility score, Score the patient burden. , , Assign weights; select the strategy with the highest utility score. As a current intervention strategy, the intervention strategy includes at least one or more combinations of medication adherence improvement, exercise prescription adjustment, dietary structure adjustment, sleep optimization, environmental exposure control, and follow-up visit reminders.

8. The dynamic monitoring and intervention management system for general practitioners with chronic diseases according to claim 1, characterized in that, The execution and feedback acquisition unit includes: a strategy distribution module, used to distribute intervention strategies. Convert to a list of executable tasks for the patient. and for each task Allocate execution time windows and reminder methods; execution record module, used to collect data on patient task completion. Timestamps, completion status, and subjective feedback; a compliance calculation module for calculating task compliance rate. ,in The function is an indicator; the execution and feedback acquisition unit further calculates the intervention effect index. ,in This is a quantified value of risk before intervention. This is the post-intervention risk quantification value; the compliance calculation module will... and Associated storage is used for subsequent strategy optimization.

9. The dynamic monitoring and intervention management system for general practitioners of chronic diseases according to claim 1, characterized in that, The closed-loop optimization unit updates the individualized baseline and policy parameters based on execution feedback data, including: updating the ITCN parameters and rhythm terms in the individualized baseline modeling unit. Seasonal items To reflect the new steady state after intervention; update the weights in the risk quantification and classification units. , , , With dynamic threshold , , To reflect changes in individual risk structure; update the strategy knowledge graph K and utility score weights in the adaptive intervention strategy generation unit. , , This reflects the true efficacy and burden of the intervention; the closed-loop optimization unit adopts an incremental learning strategy, updating only a subset of parameters related to feedback, in order to improve learning efficiency while ensuring stability.

10. The dynamic monitoring and intervention management system for general practitioners with chronic diseases according to claim 1, characterized in that, The security and compliance control unit includes: a data acquisition minimization module, used to collect only the minimum dataset necessary for monitoring and intervention; a data desensitization and anonymization module, used to desensitize identity information and sensitive data, generating anonymized identifiers; a transmission encryption module, used to encrypt data transmission and verify integrity using national cryptographic algorithms SM2 / SM3 / SM4; an access control module, used to perform access control based on roles and the principle of least privilege; an audit trail module, used to perform end-to-end auditing of data access and operations; and a data retention and destruction strategy module, used to retain data and destroy it upon expiration according to regulatory requirements. This security and compliance control unit spans the entire lifecycle of data acquisition, transmission, storage, use, and sharing, ensuring patient privacy and data security.

Images