Multi-dimensional data-driven dynamic intervention method and system for individualized rehabilitation of chronic obstructive pulmonary disease
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- BEIJING CHAOYANG HOSPITAL CAPITAL MEDICAL UNIVERSITY
- Filing Date
- 2026-03-10
- Publication Date
- 2026-06-12
Smart Images
Figure CN122201594A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of digital medical technology, specifically to a multi-dimensional data-driven method and system for dynamic intervention in individualized rehabilitation of COPD. Background Technology
[0002] Pulmonary rehabilitation has clear benefits for patients with stable COPD, but in medical practice, due to problems such as shortage of medical resources (medical institutions lack personnel, lack of pulmonary rehabilitation professional skills, and lack of time for medical staff) and low patient willingness (cannot receive treatment at home, and incur care and transportation costs), it has long been in a state of low implementation rate, low standardization rate, and poor compliance.
[0003] With the widespread adoption of mobile internet, wearable devices, and smart home IoT, digital therapy (DTx), as a novel behavioral intervention, has been explored in the field of COPD pulmonary rehabilitation. Digital healthcare products, such as the "myCOPD" and "KaiaCOPD" apps, offer functions like remote inhaler usage instruction, symptom and health data recording, medication reminders and plans, pulmonary rehabilitation courses, health education and knowledge bases, personalized action plans, remote monitoring and data sharing, community and support, and report export and printing. However, these products generally suffer from low user acceptance, declining user activity after initial downloads, insufficient long-term adherence, weak self-efficacy, and incomplete follow-up and feedback mechanisms, making it difficult to sustain interventions and impacting rehabilitation outcomes. Currently, some scholars have proposed integrating behavioral theory models into digital therapy to address the core pain points of current digital rehabilitation products through scientific behavioral intervention logic, thereby improving long-term patient engagement and rehabilitation adherence.
[0004] However, the integration of current behavioral theory models with digital COPD therapies is still in the exploratory stage and has not yet formed a mature and replicable application system. Some products that attempt to incorporate theoretical models suffer from a disconnect between theory and practice, simply adding incentive functions without deeply considering the characteristics of COPD patients (such as a high proportion of elderly patients, weak digital operation skills, and susceptibility to negative emotions such as anxiety and depression) and their rehabilitation needs (such as long-term lung function training, respiratory muscle exercises, and prevention of acute exacerbations) for functional adaptation. Furthermore, significant individual differences exist in the behavioral characteristics and psychological states of different patients, making it difficult for a single theoretical model to cover the needs of the entire population, and the adaptability and accuracy of multi-model integration are insufficient. In addition, existing integration solutions lack tracking and quantitative evaluation of intervention effects, resulting in a lack of scientific data support for product iteration.
[0005] Meanwhile, the rehabilitation process for COPD patients is long-term and complex, requiring a multi-dimensional approach encompassing medication, breathing exercises, exercise rehabilitation, dietary management, and psychological intervention. Existing digital therapies integrating behavioral theory often focus on single rehabilitation behaviors (such as medication reminders and training check-ins), lacking systematic coverage of the entire rehabilitation process. Furthermore, the generally low digital literacy among elderly patients, coupled with complex interfaces and obscure theoretical guidance in some products, further reduces user acceptance and engagement, failing to fundamentally address the core issue of "declining activity after download." Therefore, achieving deep integration of behavioral theory models with COPD digital therapies, optimizing product functionality based on individual patient characteristics and disease needs, and constructing a full-cycle, personalized, and closed-loop digital rehabilitation intervention system while improving product usability and universality have become key breakthrough directions for the development of COPD pulmonary rehabilitation digital therapies. This is also the core path to extend pulmonary rehabilitation from "medical scenarios" to "home scenarios," improving overall implementation rates and rehabilitation outcomes. Summary of the Invention
[0006] The embodiments of the present invention provide a multi-dimensional data-driven method and system for dynamic intervention in individualized rehabilitation of COPD, which can address the problems of poor patient compliance, static intervention plans, and lack of behavioral theory guidance in existing digital rehabilitation for COPD.
[0007] To achieve the above objectives, the embodiments of the present invention adopt the following technical solutions:
[0008] In a first aspect, embodiments of the present invention provide a multidimensional data-driven method for dynamic intervention in individualized rehabilitation of COPD, comprising: continuously collecting multimodal raw data from a user's terminal; cleaning and edge computing the raw data to obtain desensitized data features; inputting the desensitized data features into a preset behavioral stage classification model; and outputting the user's current pulmonary rehabilitation cognitive stage label through multidimensional feature mapping; wherein the behavioral stage classification model is constructed based on a cross-theoretical model (TTM), and the cognitive stage label includes at least the pre-intention stage, intention stage, preparation stage, action stage, and maintenance stage; performing feature fusion and quantification calculation based on the desensitized data features; calculating a relapse risk index in conjunction with the user's historical baseline data; and combining the cognitive stage... The cognitive stage tags, using a pre-defined Health Behavior Process Orientation (HAPA) logic, calculate the user's motivation intensity and execution will value. The cognitive stage tags, relapse risk index, motivation intensity, and execution will value are used as input vectors to perform node matching within a pre-defined intervention rule strategy tree to generate individualized intervention strategies. This intervention rule strategy tree is constructed based on the MAPR model and includes multiple strategy nodes corresponding to motivation enhancement, ability support, prompting strategies, and reward feedback. Based on the matched strategy nodes, corresponding digital lung rehabilitation intervention content is generated and pushed to the user's terminal. Simultaneously, the user's execution feedback data on the intervention content is continuously monitored, and the cognitive stage tags are dynamically updated based on this data.
[0009] Furthermore, the continuous collection of multimodal raw data from the user based on the user terminal includes any one or any combination of the following: collecting the user's audio stream data through the microphone module of the smart terminal device, the audio stream data being used to analyze the user's voice characteristics; collecting the user's step count and body posture data through the acceleration module of the smart terminal device; collecting the user's self-reported respiratory symptoms, post-exercise fatigue, and medication check-in data through the APP questionnaire of the smart terminal device; collecting the user's vital signs data through the heart rate sensor and arterial blood oxygen saturation sensor of the wearable device; and collecting the user's sleep duration and deep / light sleep data through the sleep monitoring module of the wearable device or smart home device.
[0010] Furthermore, the node matching in the preset intervention rule strategy tree includes: if the cognitive stage label is pre-intention period, then matching a motivation enhancement strategy node to generate disease hazard popular science content; or, if the cognitive stage label is action period and the relapse risk index is higher than a preset threshold, then matching a capability support and prompt strategy node to generate high-frequency medication reminders, breathing training and exercise training guidance content.
[0011] Furthermore, the step of cleaning and edge computing the original data to obtain desensitized data features includes: calling the local computing resources of the user terminal to extract time-domain or frequency-domain features from the original data to obtain target features including speech rate features, pitch features, motion amplitude features, and physiological parameter fluctuation features; after extracting the target features, triggering an erase command in the local storage space to destroy the corresponding original data; and encapsulating the target features into an encrypted data stream and reporting it to the cloud server.
[0012] Furthermore, the method also includes: performing timestamp alignment and numerical normalization on the received multi-source de-identified data features; using a preset weighted fusion operator or feature splicing matrix to synthesize physiological data features and behavioral data features into a high-dimensional user state vector; acquiring the user's historical baseline data stored in the cloud, and performing difference calculation or trend fitting on the high-dimensional user state vector and the baseline data to calculate the user's risk deviation value.
[0013] Furthermore, the method also includes: inputting the risk deviation value and the user's compliance assessment results into the intervention rule strategy tree to match the corresponding intervention instruction set; the intervention instruction set includes: rehabilitation training instructions, cognitive behavioral therapy (CBT) intervention instructions, dietary recommendation instructions, behavioral interaction guidance instructions, and smart home device control instructions; after pushing the intervention content to the user terminal, obtaining the user's completion rate data for the intervention content; the cloud evaluates the effectiveness of the intervention based on the completion rate data and real-time desensitized feature data, and adjusts the matching weights in the intervention rule strategy tree according to preset rules to generate an updated individualized intervention strategy, while generating an assessment report and feeding it back to the user terminal.
[0014] Secondly, embodiments of the present invention provide a multidimensional data-driven dynamic intervention system for individualized rehabilitation of COPD, comprising: a data acquisition module for continuously collecting multimodal raw data from a user based on a user terminal; a data desensitization module for cleaning and edge computing the raw data to obtain desensitized data features; a cognitive stage module for inputting the desensitized data features into a preset behavioral stage classification model, and outputting the user's current lung rehabilitation cognitive stage label through multidimensional feature mapping; the behavioral stage classification model is constructed based on the Transtheoretical Model (TTM), and the cognitive stage label includes at least the pre-intention stage, intention stage, preparation stage, action stage, and maintenance stage; and a relapse risk module for performing feature fusion and quantitative calculation based on the desensitized data features, and calculating a relapse risk index in conjunction with the user's historical baseline data. The system combines the cognitive stage tags with a preset Health Behavior Process Orientation (HAPA) logic to calculate the user's motivation intensity and execution will value. A strategy matching module uses the cognitive stage tags, relapse risk index, motivation intensity, and execution will value as input vectors to perform node matching within a preset intervention rule strategy tree to generate individualized intervention strategies. The intervention rule strategy tree is constructed based on the MAPR model and includes multiple strategy nodes corresponding to motivation enhancement, ability support, prompting strategies, and reward feedback. An intervention push module generates corresponding digital lung rehabilitation intervention content based on the matched strategy nodes and pushes it to the user terminal. Simultaneously, it continuously monitors the user's execution feedback data on the intervention content and dynamically updates the cognitive stage tags based on the execution feedback data.
[0015] Furthermore, the data acquisition module is also used to collect the user's audio stream data through the microphone module of the smart terminal device, and the audio stream data is used to analyze the user's voice characteristics; to collect the user's step count and body posture data through the acceleration module of the smart terminal device; to collect the user's self-reported respiratory symptoms, post-exercise fatigue, and medication check-in data through the APP questionnaire of the smart terminal device; to collect the user's vital signs data through the heart rate sensor and arterial blood oxygen saturation sensor of the wearable device; and to collect the user's sleep duration and deep sleep data through the sleep monitoring module of the wearable device or smart home device.
[0016] Furthermore, the strategy matching module is also used to match motivation enhancement strategy nodes to generate disease hazard popular science content if the cognitive stage label is pre-intention period; or, if the cognitive stage label is action period and the relapse risk index is higher than a preset threshold, match ability support and prompt strategy nodes to generate high-frequency medication reminders, breathing training and exercise training guidance content.
[0017] Furthermore, the data desensitization module includes: a feature extraction submodule, used to call the local computing resources of the user terminal to extract time-domain or frequency-domain features from the original data to obtain target features including speech rate features, pitch features, motion amplitude features, and physiological parameter fluctuation features; a data destruction submodule, used to trigger an erase command in the local storage space after the target features are extracted to destroy the corresponding original data; and a data encapsulation submodule, used to encapsulate the target features into an encrypted data stream and report it to the cloud server.
[0018] Furthermore, the system also includes: a data processing module for performing timestamp alignment and numerical normalization on the received multi-source de-identified data features; a feature fusion module for using a preset weighted fusion operator or feature splicing matrix to synthesize physiological data features and behavioral data features into a high-dimensional user state vector; and a risk deviation module for acquiring the user's historical baseline data stored in the cloud, performing difference calculations or trend fitting between the high-dimensional user state vector and the baseline data to calculate the user's risk deviation value.
[0019] Furthermore, the system also includes: a strategy matching module, which is further used to input the risk deviation value and the user's compliance assessment results into the intervention rule strategy tree to match the corresponding intervention instruction set; the intervention instruction set includes: rehabilitation training instructions, cognitive behavioral therapy (CBT) intervention instructions, dietary recommendation instructions, behavioral interaction guidance instructions, and smart home device control instructions; an intervention completion module, which is used to obtain the user's completion rate data of the intervention content after pushing the intervention content to the user terminal; and an intervention update module, which is used to evaluate the effectiveness of the intervention in the cloud based on the completion rate data and real-time desensitized feature data, and adjust the matching weight in the intervention rule strategy tree according to preset rules to generate an updated individualized intervention strategy, while generating an evaluation report and feeding it back to the user terminal.
[0020] Compared with the prior art, the beneficial effects of the present invention include:
[0021] Compared with existing pulmonary rehabilitation technologies for elderly COPD patients, this invention achieves a comprehensive breakthrough in addressing clinical pain points, and has significant advantages in intervention effectiveness, implementation feasibility, user compliance, and large-scale promotion.
[0022] 1. Current pulmonary rehabilitation for elderly patients with COPD heavily relies on one-on-one guidance from professional medical staff. However, limitations such as staff shortages in primary healthcare institutions and the concentration of high-quality rehabilitation resources in tertiary hospitals prevent most patients from accessing standardized rehabilitation services. This invention reconstructs the rehabilitation service model through digital technology, deeply integrating evidence-based COPD rehabilitation intervention strategies (breathing training, exercise rehabilitation, health education, etc.) with three major behavioral change theories: the Transtheoretical Model (TTM), the Health Behavior Process Orientation Theory (HAPA), and the MAPR model. It achieves automated and standardized rehabilitation interventions through apps, smart terminals, and wearable devices. The core rehabilitation process can be completed without the need for full-time professional medical staff, breaking down limitations of manpower, location, and time. This allows elderly patients to access standardized and consistent pulmonary rehabilitation services without leaving their homes, effectively addressing the poor accessibility of rehabilitation services due to a shortage of professional medical staff and achieving the universalization and downward flow of rehabilitation resources.
[0023] 2. Existing offline rehabilitation programs suffer from inconsistent implementation procedures and arbitrary execution. Furthermore, elderly patients often face mobility issues and high medical costs, resulting in extremely low implementation rates and difficulties in ensuring the standardization of program implementation. This invention constructs a digital intervention system based on behavioral theory, breaking down rehabilitation interventions into standardized digital modules tailored to elderly patients. Through hardware and software collaboration, a closed-loop management system is achieved throughout the entire process: wearable devices monitor patients' vital signs (blood oxygen, heart rate, activity level) in real time; an app dynamically pushes personalized rehabilitation plans based on behavioral theory; and smart terminals automatically record intervention execution. This fundamentally avoids the biases inherent in manual execution, ensuring that rehabilitation intervention steps, frequency, and intensity strictly adhere to evidence-based medicine guidelines. It achieves standardized and traceable implementation of pulmonary rehabilitation for elderly patients with COPD, significantly improving the implementation rate and standardization of rehabilitation programs.
[0024] 3. Elderly COPD patients commonly experience low self-efficacy, insufficient understanding, and difficulty in adhering to rehabilitation behaviors. Traditional rehabilitation models suffer from drawbacks such as cumbersome procedures, lack of interaction, and absence of sustained motivation, leading to poor patient participation and high dropout rates. This invention designs a dynamic intervention mechanism based on behavioral change theory, organically integrating three major behavioral theories—the Transtheoretical Model (TTM), the Health Behavior Process Orientation (HAPA) model, and the MAPR model—with evidence-based COPD rehabilitation intervention strategies. This multi-dimensional approach stimulates patients' willingness to rehabilitate, addresses the core issues of low self-efficacy and poor adherence among elderly patients, improves patient participation, sustained adherence, and rehabilitation persistence, and ensures the long-term effectiveness of rehabilitation interventions.
[0025] 4. Existing digital healthcare products are mostly geared towards younger users, with complex operations and cumbersome interfaces, and lack design adaptations for elderly COPD patients, resulting in high barriers to entry and resistance among them. This invention focuses on the physiology and usage habits of the elderly, optimizing both hardware and software for age-friendliness: the app features large fonts, minimalist interaction, and voice navigation; wearable devices offer convenient operation; and smart terminals provide one-click help and abnormal warning functions. Simultaneously, personalized push notifications based on behavioral theory avoid information overload. This significantly reduces the difficulty of use for elderly patients, adapting to their vision, operational abilities, and other physiological characteristics, solving the problem of poor usability in digital products, improving patient willingness and experience, and laying a user foundation for long-term rehabilitation intervention.
[0026] 5. Current pulmonary rehabilitation assessments largely rely on regular offline follow-ups, resulting in data lag and untimely feedback, making it difficult to dynamically adjust treatment plans based on patient progress. This invention utilizes wearable devices, an app, and smart terminals to achieve real-time data collection and intelligent analysis. It monitors multi-dimensional data such as patient mobility and symptom scores in real time, dynamically optimizing rehabilitation plans based on behavioral theory. Simultaneously, it comprehensively records the patient's rehabilitation behavior trajectory, providing data support for individualized rehabilitation adjustments. This enables precise and personalized management of pulmonary rehabilitation for elderly COPD patients, allowing for timely adjustments to intervention strategies to adapt to the patient's rehabilitation progress. Compared to traditional static rehabilitation programs, it is expected to further improve patient symptoms, enhance mobility, reduce the risk of acute exacerbations, and improve clinical rehabilitation outcomes.
[0027] 6. Traditional offline pulmonary rehabilitation programs are limited by venue, manpower, and cost, making large-scale promotion difficult; existing digital rehabilitation products mostly lack theoretical support and standardized design, resulting in insufficient adaptability and sustainability. This invention integrates behavioral theory and evidence-based medicine to construct a replicable and scalable digital rehabilitation system, adaptable to multiple scenarios such as tertiary hospitals, primary community hospitals, and home-based elderly care, with low hardware and software deployment costs and convenient operation and maintenance. It aligns with the development trends of hierarchical diagnosis and treatment, home-based rehabilitation, and digital transformation in my country's medical system, and can be quickly implemented in medical institutions at all levels, elderly care institutions, and home settings, providing a feasible technical solution for the large-scale and routine implementation of pulmonary rehabilitation for elderly patients with COPD, possessing significant social and application value. Attached Figure Description
[0028] Figure 1 This is a flowchart illustrating the multidimensional data-driven dynamic intervention method for individualized rehabilitation of COPD in this specific embodiment.
[0029] Figure 2 This is another flowchart illustrating the multidimensional data-driven dynamic intervention method for individualized rehabilitation of COPD in this specific implementation.
[0030] Figure 3This is a schematic diagram of the structure of the multi-dimensional data-driven dynamic intervention system for individualized rehabilitation of COPD in this specific embodiment;
[0031] Figure 4 This is a schematic diagram of the data desensitization module in this specific embodiment;
[0032] Figure 5 This is another structural diagram of the multidimensional data-driven dynamic intervention system for individualized rehabilitation of COPD in this specific embodiment. Detailed Implementation
[0033] The present invention will now be described in further detail with reference to the accompanying drawings:
[0034] refer to Figure 1 As shown in this specific embodiment, the multidimensional data-driven dynamic intervention method for individualized rehabilitation of COPD includes:
[0035] 101. Based on the user terminal, continuously collect multimodal raw data of the user, clean the raw data and perform edge computing to obtain de-identified data features.
[0036] The user terminal includes user-associated smart terminal devices, wearable devices, and smart home devices. Specifically, smart terminal devices may include the user's smartphone or smart tablet. Wearable devices may include the user's smartwatch, smart bracelet, smart ring, smart vest, etc. Smart home devices may include smart curtains, smart thermometers, smart hygrometers, smart decibel meters, smart lights, smart mattresses, etc.
[0037] In this embodiment of the invention, the microphone module of a smart terminal device can collect the user's audio stream data, which is used to analyze the user's voice characteristics; the accelerometer module of the smart terminal device can collect the user's step count and body posture data; the APP questionnaire of the smart terminal device can collect the user's self-reported respiratory symptoms, post-exercise fatigue, and medication record data; the heart rate sensor and arterial blood oxygen saturation sensor of a wearable device can collect the user's vital signs data; and the sleep monitoring module of a wearable device or smart home device can collect the user's sleep duration and deep / light sleep data. The user's voice characteristics may include semantic, speech rate, and tone features.
[0038] 102. Input the desensitized data features into a preset behavioral stage classification model, and output the user's current lung rehabilitation cognitive stage label through multidimensional feature mapping.
[0039] The behavioral stage classification model is constructed based on the Transtheoretical Model (TTM), and the cognitive stage labels include at least the pre-intention stage, intention stage, preparation stage, action stage, and maintenance stage.
[0040] 103. Based on the desensitized data features, perform feature fusion and quantification calculations, and calculate the relapse risk index by combining the user's historical baseline data; and combine the cognitive stage tags, and use the preset Health Behavior Process Orientation (HAPA) logic to calculate the user's motivation intensity value and execution will value.
[0041] 104. Using the cognitive stage label, relapse risk index, motivation intensity value, and execution will value as input vectors, perform node matching in a preset intervention rule strategy tree to generate an individualized intervention strategy.
[0042] The intervention rule strategy tree is constructed based on the MAPR model and contains multiple strategy nodes corresponding to motivation enhancement, capability support, prompting strategies, and reward feedback, respectively.
[0043] 105. Based on the matched strategy node, generate corresponding digital lung rehabilitation intervention content and push it to the user terminal; at the same time, continuously monitor the user's execution feedback data on the intervention content, and dynamically update the cognitive stage label based on the execution feedback data.
[0044] In this embodiment of the invention, the system platform adopts a layered architecture, with each layer interacting through a defined data flow to collaboratively complete the entire process from data collection to treatment intervention feedback. In this embodiment, the layered architecture enables "local feature extraction and cloud-based model inference," ensuring the privacy of the user's original data while effectively analyzing the user's psychological state.
[0045] Compared with existing technologies, the embodiments of this invention can collect data through user terminals, continuously acquiring objective and authentic user data. This avoids the problems of inaccurate and untimely data information when users provide subjective feedback, which are common in existing technologies. Furthermore, by collecting multi-dimensional data for comprehensive evaluation and analysis, such as user behavior data, user physiological state, and user sleep quality data, the accuracy and timeliness of user psychological state assessment can be effectively improved. Moreover, through real-time data collection, psychological state assessment, and digital therapy intervention, immediate risk feedback and intervention treatment can be achieved, seizing the "golden window" of therapeutic intervention. This effectively enhances the intervention effect on user psychological state, increases clinical applicability and user trust, and constructs a closed-loop system architecture of IoT sensing, data analysis, and digital intervention.
[0046] refer to Figure 2 As shown in this specific embodiment, another multidimensional data-driven dynamic intervention method for individualized rehabilitation of COPD includes:
[0047] 201. Continuously collect multimodal raw data from users based on user terminals.
[0048] The user terminal includes user-associated smart terminal devices, wearable devices, and smart home devices. Specifically, smart terminal devices may include the user's smartphone or smart tablet. Wearable devices may include the user's smartwatch, smart bracelet, smart ring, smart vest, etc. Smart home devices may include smart curtains, smart thermometers, smart hygrometers, smart decibel meters, smart lights, smart mattresses, etc.
[0049] In this embodiment of the invention, the microphone module of a smart terminal device can collect the user's audio stream data, which is used to analyze the user's voice characteristics; the accelerometer module of the smart terminal device can collect the user's step count and body posture data; the APP questionnaire of the smart terminal device can collect the user's self-reported respiratory symptoms, post-exercise fatigue, and medication record data; the heart rate sensor and arterial blood oxygen saturation sensor of a wearable device can collect the user's vital signs data; and the sleep monitoring module of a wearable device or smart home device can collect the user's sleep duration and deep / light sleep data. The user's voice characteristics may include semantic, speech rate, and tone features.
[0050] 202. Call the local computing resources of the user terminal to extract time-domain or frequency-domain features from the original data to obtain target features including speech rate features, pitch features, motion amplitude features and physiological parameter fluctuation features.
[0051] 203. After extracting the target features, trigger the erase command in the local storage space to destroy the corresponding original data.
[0052] In this embodiment of the invention, by promptly erasing the original user data after local feature extraction, user privacy can be effectively protected, eliminating users' privacy concerns and stigma, improving compliance, and thus enabling long-term monitoring of users' psychological state.
[0053] 204. Encapsulate the target features into an encrypted data stream and report it to the cloud server.
[0054] 205. Input the desensitized data features into a preset behavioral stage classification model, and output the user's current lung rehabilitation cognitive stage label through multidimensional feature mapping.
[0055] The behavioral stage classification model is constructed based on the Transtheoretical Model (TTM), and the cognitive stage labels include at least the pre-intention stage, intention stage, preparation stage, action stage, and maintenance stage.
[0056] 206. Based on the desensitized data features, perform feature fusion and quantification calculations, and calculate the relapse risk index by combining the user's historical baseline data; and combine the cognitive stage tags, and use the preset Health Behavior Process Orientation (HAPA) logic to calculate the user's motivation intensity value and execution will value.
[0057] In specific embodiments of the present invention, the received multi-source de-identified data features can be timestamped and normalized; physiological data features and behavioral data features can be synthesized into a high-dimensional user state vector using a preset weighted fusion operator or feature splicing matrix; the user's historical baseline data stored in the cloud can be obtained, and the high-dimensional user state vector and the baseline data can be subjected to difference calculation or trend fitting to calculate the user's risk deviation value.
[0058] In this embodiment of the invention, a mapping logic engine based on TTM, HAPA, and MAPR theories can be pre-configured to construct a multi-dimensional theoretical mapping matrix. Specifically, the TTM model can be transformed into the system's timing control logic, defining the time thresholds and task load parameters for the three stages of adaptation, reinforcement, and consolidation; the HAPA model can be transformed into node triggering rules, automatically switching the functional priorities of risk perception, execution monitoring, and social support at different stages; and the MAPR model can be transformed into an interactive control strategy, configuring corresponding motivation enhancement algorithms, capability support modules, prompt instruction sets, and reward engines.
[0059] 207. Using the cognitive stage label, relapse risk index, motivation intensity value, and execution will value as input vectors, perform node matching in a pre-set intervention rule strategy tree to generate an individualized intervention strategy; the intervention rule strategy tree is constructed based on the MAPR model and contains multiple strategy nodes corresponding to motivation enhancement, ability support, prompting strategy, and reward feedback, respectively.
[0060] 208. If the cognitive stage label is pre-intention stage, then match the motivation enhancement strategy node to generate disease hazard popular science content; or, if the cognitive stage label is action stage and the relapse risk index is higher than the preset threshold, then match the ability support and prompt strategy node to generate high-frequency medication reminders, breathing training and exercise training guidance content.
[0061] 209. Input the risk deviation value and the user's compliance assessment result into the intervention rule strategy tree and match the corresponding intervention instruction set.
[0062] The intervention instruction set includes: rehabilitation training instructions, cognitive behavioral therapy (CBT) intervention instructions, dietary recommendation instructions, behavioral interaction guidance instructions, and smart home device control instructions.
[0063] 210. After pushing the intervention content to the user terminal, obtain the user's completion rate data for the intervention content.
[0064] 211. The cloud evaluates the intervention effectiveness based on the completion data and real-time de-identified feature data, adjusts the matching weights in the intervention rule strategy tree according to preset rules, generates an updated personalized intervention strategy, and simultaneously generates an evaluation report and feeds it back to the user terminal.
[0065] In an embodiment of the present invention, the system functions are modularly integrated, integrating core functional modules such as health assessment, rehabilitation training, risk warning, smoking cessation intervention, and dietary nutrition. Among them, the rehabilitation training module collects physiological and exercise data in real time through the App terminal, and adjusts the interaction logic by calling the corresponding MAPR strategy according to the current TTM stage.
[0066] Exemplarily, the system functions may include: (1) Rehabilitation plan: personalized plan generation / viewing, daily medication (inhaled and oral medications) check-in / comment, daily respiratory muscle training / resistance exercise (follow-along video and check-in) check-in, daily aerobic exercise respiratory muscle training (exercise and heart rate linkage) check-in, patient education learning (micro-course), disease state assessment (questionnaire); (2) Health consultation: intelligent agent communication (assistance), doctor-patient communication; (3) Health monitoring: wearable device data viewing, periodic rehabilitation weekly and monthly reports; (4) Risk monitoring: health risk identification and warning, first aid guidance and education, compliance risk identification and warning; (5) Result sharing: generation and sharing of check-in achievement posters.
[0067] In an embodiment of the present invention, a dynamic path can be generated. During the adaptation period (for example, T1 ≤ 2 weeks), the system sets a relatively low initial training intensity through the "lowering the action threshold" algorithm, and matches high-frequency "timed push reminders" and "progress visualization" instructions to reduce the user's start-up barriers. During the strengthening period (for example, 2 weeks < T2 ≤ 6 weeks), the system automatically raises the training load threshold and activates the "incentive feedback" and "real-time data comparison" modules to strengthen the user's behavior. During the consolidation period (for example, T3 > 6 weeks), the system switches to the "achievement system" and "effect tracking" strategies, and promotes habit solidification through the driving force maintenance algorithm.
[0068] In an embodiment of the present invention, under the rehabilitation training function, the system can also receive user feedback in real time through the "program self-adjustment" sub-module. If the user's physiological parameters are abnormal or the compliance decreases, the system automatically reverts to the previous intervention stage or regenerates a response plan.
[0069] In an embodiment of the present invention, based on the user's historical analysis data, the personalized baseline of the user is dynamically adjusted and the intervention rule strategy is dynamically optimized, fully considering individual differences, and a customized intervention plan can be provided for the user to achieve personalized evaluation and intervention of "one size fits one person".
[0070] In this embodiment of the invention, changes in user data after intervention are used as indicators for efficacy evaluation, enabling an adaptive optimization of the feedback mechanism for intervention strategies. This allows for the construction of a closed-loop system architecture encompassing IoT sensing, data analysis, and digital intervention. In this embodiment, to meet the data localization requirements of medical institutions, the entire platform system can also be deployed within the hospital's private cloud, integrated with the hospital's HIS (Hospital Information System), serving as a clinical support tool, and connected to smart devices within wards.
[0071] Compared with existing technologies, the embodiments of this invention can collect data through user terminals, continuously acquiring objective and authentic user data. This avoids the problems of inaccurate and untimely data information when users provide subjective feedback, which are common in existing technologies. Furthermore, by collecting multi-dimensional data for comprehensive evaluation and analysis, such as user behavior data, user physiological state, and user sleep quality data, the accuracy and timeliness of user psychological state assessment can be effectively improved. Moreover, through real-time data collection, psychological state assessment, and digital therapy intervention, immediate risk feedback and intervention treatment can be achieved, seizing the "golden window" of therapeutic intervention. This effectively enhances the intervention effect on user psychological state, increases clinical applicability and user trust, and constructs a closed-loop system architecture of IoT sensing, data analysis, and digital intervention.
[0072] Compared with existing pulmonary rehabilitation technologies for elderly COPD patients, this invention achieves a comprehensive breakthrough in addressing clinical pain points, demonstrating significant advantages in intervention effectiveness, implementation feasibility, user compliance, and scalability. Current pulmonary rehabilitation for elderly COPD patients heavily relies on one-on-one guidance from professional medical staff. Limited by issues such as staff shortages in primary healthcare institutions and the concentration of high-quality rehabilitation resources in tertiary hospitals, many patients are unable to access standardized rehabilitation services. This invention reconstructs the rehabilitation service model through digital technology, deeply integrating evidence-based COPD rehabilitation intervention strategies (breathing training, exercise rehabilitation, health education, etc.) with three major behavioral change theories: Transtheoretical Model (TTM), Health Behavior Process Orientation Theory (HAPA), and MAPR model. It achieves automated and standardized rehabilitation interventions through apps, smart terminals, and wearable devices. The core rehabilitation process can be completed without the full intervention of professional medical staff, breaking down limitations of manpower, location, and time. This allows elderly patients to access homogeneous and standardized pulmonary rehabilitation services without leaving their homes, effectively solving the problem of poor accessibility to rehabilitation services due to a shortage of professional medical staff and achieving the universalization and downward flow of rehabilitation resources.
[0073] Compared to existing offline rehabilitation programs, which suffer from inconsistent implementation processes and arbitrary execution, and are hampered by the mobility limitations and high medical costs faced by elderly patients, resulting in extremely low rehabilitation implementation rates and difficulties in ensuring the standardization of program implementation, this invention constructs a digital intervention system based on behavioral theory. It breaks down rehabilitation interventions into standardized digital modules tailored to elderly patients, achieving closed-loop management throughout the entire process through hardware and software collaboration: wearable devices monitor patients' vital signs (blood oxygen, heart rate, activity level) in real time; an app dynamically pushes personalized rehabilitation plans based on behavioral theory; and smart terminals automatically record intervention execution, fundamentally avoiding deviations from manual execution. This ensures that rehabilitation intervention steps, frequency, and intensity strictly adhere to evidence-based medicine guidelines, achieving standardized and traceable implementation of pulmonary rehabilitation for elderly patients with COPD, significantly improving the implementation rate and standardization of rehabilitation programs.
[0074] Meanwhile, elderly COPD patients generally suffer from low self-efficacy, insufficient cognition, and difficulty in adhering to rehabilitation behaviors. Furthermore, traditional rehabilitation models are cumbersome, lack interaction, and fail to provide sustained motivation, resulting in poor patient participation and high dropout rates in existing programs. This invention, based on behavioral change theory, designs a dynamic intervention mechanism that organically integrates three major behavioral theories—the Transtheoretical Model (TTM), the Health Behavior Process Orientation (HAPA) model, and the MAPR model—with evidence-based COPD rehabilitation intervention strategies. This multi-dimensional approach stimulates patients' willingness to rehabilitate, addresses the core issues of low self-efficacy and poor adherence among elderly patients, improves patient participation, sustained adherence, and rehabilitation persistence, and ensures the long-term effectiveness of the rehabilitation intervention.
[0075] Compared to existing digital healthcare products, which are mostly geared towards younger users, are complex to operate, have cumbersome interfaces, and lack designs adapted for elderly COPD patients, resulting in high barriers to entry and resistance among them, this invention focuses on the physiology and usage habits of the elderly, optimizing both hardware and software for age-friendliness: the app features large fonts, minimalist interaction, and voice navigation; wearable devices offer convenient operation; and smart terminals provide one-click help and abnormal warning functions. Furthermore, personalized push notifications based on behavioral theory prevent information overload. This significantly reduces the difficulty of use for elderly patients, adapts to their vision, operational abilities, and other physiological characteristics, solves the problem of poor usability in digital products, improves patient willingness and experience, and lays a user foundation for long-term rehabilitation intervention.
[0076] Compared to existing pulmonary rehabilitation assessments that rely heavily on regular offline follow-ups, which suffer from data lag and untimely feedback, making it difficult to dynamically adjust treatment plans based on patient progress, this invention utilizes wearable devices, an app, and smart terminals to achieve real-time data collection and intelligent analysis. It monitors multi-dimensional data such as patient mobility and symptom scores in real time, dynamically optimizing rehabilitation plans based on behavioral theory. Simultaneously, it comprehensively records the patient's rehabilitation behavior trajectory, providing data support for individualized rehabilitation adjustments. This enables precise and personalized management of pulmonary rehabilitation for elderly COPD patients, allowing for timely adjustments to intervention strategies to adapt to the patient's rehabilitation progress. Compared to traditional static rehabilitation programs, it is expected to further improve patient symptoms, enhance mobility, reduce the risk of acute exacerbations, and improve clinical rehabilitation outcomes.
[0077] Compared to traditional offline pulmonary rehabilitation programs, which are limited by venue, manpower, and cost, making large-scale promotion difficult, existing digital rehabilitation products often lack theoretical support and standardized design, resulting in insufficient adaptability and sustainability. This invention integrates behavioral theory and evidence-based medicine to construct a replicable and scalable digital rehabilitation system, adaptable to various scenarios including tertiary hospitals, primary community hospitals, and home-based elderly care. It also boasts low hardware and software deployment costs and convenient operation and maintenance. Aligning with the development trends of hierarchical medical treatment, home-based rehabilitation, and digital transformation in my country's medical system, it can be rapidly implemented in medical institutions at all levels, elderly care facilities, and home settings, providing a feasible technical solution for the large-scale and routine implementation of pulmonary rehabilitation for elderly patients with COPD, and possessing significant social and application value.
[0078] refer to Figure 3 As shown in this specific embodiment, the multidimensional data-driven dynamic intervention system for individualized COPD rehabilitation includes:
[0079] The data acquisition module 31 is used to continuously collect multimodal raw data from users based on user terminals.
[0080] The data desensitization module 32 is used to clean and perform edge calculations on the original data to obtain desensitized data features.
[0081] The cognitive stage module 33 is used to input the desensitized data features into a preset behavioral stage classification model and output the user's current pulmonary rehabilitation cognitive stage label through multidimensional feature mapping. The behavioral stage classification model is constructed based on the cross-theoretical model TTM, and the cognitive stage label includes at least the pre-intention period, intention period, preparation period, action period, and maintenance period.
[0082] The relapse risk module 34 is used to perform feature fusion and quantification calculation based on the desensitized data features, calculate the relapse risk index by combining the user's historical baseline data, and calculate the user's motivation intensity value and execution will value by combining the cognitive stage label and using the preset Health Behavior Process Orientation (HAPA) logic.
[0083] The strategy matching module 35 is used to use the cognitive stage label, relapse risk index, motivation intensity value and execution will value as input vectors to perform node matching in a preset intervention rule strategy tree to generate an individualized intervention strategy; the intervention rule strategy tree is constructed based on the MAPR model and contains multiple strategy nodes corresponding to motivation enhancement, ability support, prompting strategy and reward feedback respectively.
[0084] The intervention push module 36 is used to generate corresponding digital lung rehabilitation intervention content based on the matched strategy node and push it to the user terminal; at the same time, it continuously monitors the user's execution feedback data of the intervention content and dynamically updates the cognitive stage label based on the execution feedback data.
[0085] Furthermore, the data acquisition module 31 is also used to acquire the user's audio stream data through the microphone module of the smart terminal device, the audio stream data being used to analyze the user's voice characteristics; to acquire the user's step count and body posture data through the acceleration module of the smart terminal device; to acquire the user's self-reported respiratory symptoms, post-exercise fatigue, and medication check-in data through the APP questionnaire of the smart terminal device; to acquire the user's vital signs data through the heart rate sensor and arterial blood oxygen saturation sensor of the wearable device; and to acquire the user's sleep duration and deep / light sleep data through the sleep monitoring module of the wearable device or smart home device.
[0086] Furthermore, the strategy matching module 35 is also used to match motivation enhancement strategy nodes to generate disease hazard popular science content if the cognitive stage label is pre-intention period; or, if the cognitive stage label is action period and the relapse risk index is higher than a preset threshold, match ability support and prompt strategy nodes to generate high-frequency medication reminders, breathing training and exercise training guidance content.
[0087] refer to Figure 4 As shown, the data desensitization module 32 further includes:
[0088] The feature extraction submodule 3201 is used to call the local computing resources of the user terminal to perform time-domain or frequency-domain feature extraction on the original data to obtain target features including speech rate features, pitch features, motion amplitude features and physiological parameter fluctuation features.
[0089] The data destruction submodule 3202 is used to trigger an erase command in the local storage space after the target features are extracted, thereby destroying the corresponding original data.
[0090] The data encapsulation submodule 3203 is used to encapsulate the target features into an encrypted data stream and report it to the cloud server.
[0091] refer to Figure 5As shown, the system further includes: a data processing module 41, a feature fusion module 42, an intervention completion module 43, and an intervention update module 44.
[0092] The data processing module 41 is used to perform timestamp alignment and numerical normalization on the received multi-source de-identified data features.
[0093] The feature fusion module 42 is used to synthesize physiological data features and behavioral data features into a high-dimensional user state vector by using a preset weighted fusion operator or feature splicing matrix.
[0094] The recurrence risk module 34 is also used to acquire the user's historical baseline data stored in the cloud, and perform difference calculation or trend fitting between the high-dimensional user state vector and the baseline data to calculate the user's risk deviation value.
[0095] The strategy matching module 35 is further used to input the risk deviation value and the user's compliance assessment results into the intervention rule strategy tree to match the corresponding intervention instruction set; the intervention instruction set includes: rehabilitation training instructions, cognitive behavioral therapy (CBT) intervention instructions, dietary recommendation instructions, behavioral interaction guidance instructions, and smart home device control instructions.
[0096] The intervention completion module 43 is used to obtain the user's completion rate data of the intervention content after pushing the intervention content to the user terminal.
[0097] The intervention update module 44 is used to evaluate the effectiveness of the intervention in the cloud based on the completion data and real-time desensitized feature data, adjust the matching weight in the intervention rule strategy tree according to preset rules, generate an updated individualized intervention strategy, and generate an evaluation report and feed it back to the user terminal.
[0098] The multidimensional data-driven dynamic intervention system for individualized rehabilitation of COPD provided in this specific embodiment can realize the above-mentioned method implementation method. For specific functional implementation, please refer to the description in the method embodiment, which will not be repeated here.
[0099] The above technical solution is only one embodiment of the present invention. For those skilled in the art, based on the principles disclosed in the present invention, it is easy to make various types of improvements or modifications, and not limited to the technical solutions described in the specific embodiments of the present invention. Therefore, the foregoing description is only a preferred option and is not restrictive.
Claims
1. A multidimensional data-driven method for individualized dynamic intervention in COPD rehabilitation, characterized in that, include: Based on the continuous collection of multimodal raw data from users' terminals, the raw data is cleaned and edge computing is performed to obtain desensitized data features; The desensitized data features are input into a pre-set behavioral stage classification model, and the user's current pulmonary rehabilitation cognitive stage label is output through multi-dimensional feature mapping. The behavioral stage classification model is constructed based on the Transtheoretical Model (TTM), and the cognitive stage label includes at least the pre-intention stage, intention stage, preparation stage, action stage, and maintenance stage. Based on the desensitized data features, feature fusion and quantification are performed, and the recurrence risk index is calculated by combining the user's historical baseline data. In conjunction with the aforementioned cognitive stage tags, and using the preset Health Behavior Process Orientation (HAPA) logic, the user's motivation intensity value and execution will value are calculated. The cognitive stage label, relapse risk index, motivation intensity value, and execution will value are used as input vectors to perform node matching in a pre-set intervention rule strategy tree to generate individualized intervention strategies. The intervention rule strategy tree is constructed based on the MAPR model and contains multiple strategy nodes corresponding to motivation enhancement, capability support, prompting strategies, and reward feedback, respectively. Based on the matched strategy nodes, corresponding digital lung rehabilitation intervention content is generated and pushed to the user terminal; at the same time, the user's execution feedback data on the intervention content is continuously monitored, and the cognitive stage tags are dynamically updated based on the execution feedback data.
2. The method according to claim 1, characterized in that, The continuous collection of multimodal raw data from users based on user terminals includes any one or any combination of the following: The user's audio stream data is collected through the microphone module of the smart terminal device, and the audio stream data is used to analyze the user's voice characteristics. The system collects users' step count and body posture data through the acceleration module of smart terminal devices; it collects users' self-reported respiratory symptoms, post-exercise fatigue, and medication check-in data through the questionnaire of the smart terminal device's APP; it collects users' vital signs data through the heart rate sensor and arterial blood oxygen saturation sensor of wearable devices; and it collects users' sleep duration and deep sleep data through the sleep monitoring module of wearable devices or smart home devices.
3. The method according to claim 1, characterized in that, The node matching in the pre-set intervention rule strategy tree includes: If the cognitive stage label is pre-intention stage, then a motivation enhancement strategy node is matched to generate disease hazard popular science content; or, if the cognitive stage label is action stage and the relapse risk index is higher than a preset threshold, then a capability support and prompt strategy node is matched to generate high-frequency medication reminders, breathing training and exercise training guidance content.
4. The method according to claim 1, characterized in that, The process of cleaning and edge computing the original data to obtain de-identified data features includes: The local computing resources of the user terminal are invoked to extract time-domain or frequency-domain features from the raw data to obtain target features including speech rate features, pitch features, motion amplitude features, and physiological parameter fluctuation features. After the target features are extracted, an erase command is triggered in the local storage space to destroy the corresponding original data; The target features are encapsulated into an encrypted data stream and reported to the cloud server.
5. The method according to claim 1, characterized in that, The step of performing feature fusion and quantification calculations based on the desensitized data features, and calculating the relapse risk index in conjunction with the user's historical baseline data, includes: The received multi-source de-identified data features are timestamped and their values are normalized. Using a preset weighted fusion operator or feature concatenation matrix, physiological data features and behavioral data features are synthesized into a high-dimensional user state vector; The user's historical baseline data stored in the cloud is obtained, and the high-dimensional user state vector is compared with the baseline data by difference calculation or trend fitting to calculate the user's risk deviation value.
6. The method according to claim 1, characterized in that, The corresponding digital lung rehabilitation intervention content is generated based on the matched strategy node and pushed to the user terminal; Simultaneously, continuously monitor user feedback data on the intervention content, and dynamically update the cognitive stage tags based on the feedback data, including: The risk deviation value and the user's compliance assessment results are input into the intervention rule strategy tree to match the corresponding intervention instruction set; the intervention instruction set includes: rehabilitation training instructions, cognitive behavioral therapy (CBT) intervention instructions, dietary recommendation instructions, behavioral interaction guidance instructions, and smart home device control instructions; After pushing the intervention content to the user terminal, the user's completion rate data for the intervention content is obtained; Based on the completion data and real-time de-identification feature data, the cloud platform evaluates the effectiveness of the intervention, adjusts the matching weights in the intervention rule strategy tree according to preset rules, generates an updated individualized intervention strategy, and generates an evaluation report that is fed back to the user terminal.
7. A multi-dimensional data-driven dynamic intervention system for individualized rehabilitation of COPD, characterized in that, include: The data acquisition module is used to continuously collect multimodal raw data from users based on user terminals; The data desensitization module is used to clean and perform edge calculations on the original data to obtain desensitized data features. The cognitive stage module is used to input the desensitized data features into a preset behavioral stage classification model, and output the user's current pulmonary rehabilitation cognitive stage label through multidimensional feature mapping; the behavioral stage classification model is constructed based on the Transtheoretical Model (TTM), and the cognitive stage label includes at least the pre-intention stage, intention stage, preparation stage, action stage, and maintenance stage; The relapse risk module is used to perform feature fusion and quantification calculations based on the desensitized data features, and to calculate the relapse risk index by combining the user's historical baseline data. In conjunction with the aforementioned cognitive stage tags, and using the preset Health Behavior Process Orientation (HAPA) logic, the user's motivation intensity value and execution will value are calculated. The strategy matching module is used to take the cognitive stage label, relapse risk index, motivation intensity value and execution will value as input vectors, and perform node matching in a preset intervention rule strategy tree to generate an individualized intervention strategy. The intervention rule strategy tree is constructed based on the MAPR model and contains multiple strategy nodes corresponding to motivation enhancement, capability support, prompting strategies, and reward feedback, respectively. The intervention push module is used to generate corresponding digital lung rehabilitation intervention content based on the matched strategy node and push it to the user terminal; at the same time, it continuously monitors the user's execution feedback data of the intervention content and dynamically updates the cognitive stage tag based on the execution feedback data.
8. The system according to claim 7, characterized in that, The data acquisition module is also used to acquire the user's audio stream data through the microphone module of the smart terminal device, and the audio stream data is used to analyze the user's voice characteristics. The system collects users' step count and body posture data through the acceleration module of smart terminal devices; it collects users' self-reported respiratory symptoms, post-exercise fatigue, and medication check-in data through the questionnaire of the smart terminal device's APP; it collects users' vital signs data through the heart rate sensor and arterial blood oxygen saturation sensor of wearable devices; and it collects users' sleep duration and deep sleep data through the sleep monitoring module of wearable devices or smart home devices.
9. The system according to claim 7, characterized in that, The strategy matching module is further configured to match motivation enhancement strategy nodes and generate disease hazard popular science content if the cognitive stage label is pre-intention period; or, if the cognitive stage label is action period and the relapse risk index is higher than a preset threshold, match capability support and prompt strategy nodes and generate high-frequency medication reminders, breathing training and exercise training guidance content.
10. The system according to claim 7, characterized in that, The data desensitization module includes: The feature extraction submodule is used to call the local computing resources of the user terminal to perform time-domain or frequency-domain feature extraction on the raw data to obtain target features including speech rate features, pitch features, motion amplitude features and physiological parameter fluctuation features. The data destruction submodule is used to trigger an erase command in the local storage space after the target features are extracted, thereby destroying the corresponding original data. The data encapsulation submodule is used to encapsulate the target features into an encrypted data stream and report it to the cloud server.
11. The system according to claim 7, characterized in that, The system also includes: The data processing module is used to perform timestamp alignment and numerical normalization on the features of the received multi-source de-identified data. The feature fusion module is used to synthesize physiological data features and behavioral data features into a high-dimensional user state vector using a preset weighted fusion operator or feature concatenation matrix. The recurrence risk module is also used to acquire the user's historical baseline data stored in the cloud, and perform difference calculation or trend fitting between the high-dimensional user state vector and the baseline data to calculate the user's risk deviation value.
12. The system according to claim 7, characterized in that, The system also includes: The strategy matching module is also used to input the risk deviation value and the user's compliance assessment results into the intervention rule strategy tree to match the corresponding intervention instruction set; the intervention instruction set includes: rehabilitation training instructions, cognitive behavioral therapy (CBT) intervention instructions, dietary recommendation instructions, behavioral interaction guidance instructions, and smart home device control instructions; The intervention completion module is used to obtain user completion data of the intervention content after pushing the intervention content to the user terminal. The intervention update module is used to evaluate the effectiveness of the intervention in the cloud based on the completion data and real-time de-identified feature data, adjust the matching weights in the intervention rule strategy tree according to preset rules, generate an updated individualized intervention strategy, and generate an evaluation report and feed it back to the user terminal.