Stroke patient rehabilitation conditioning system and method based on self-learning technology

Abstract

The invention discloses a stroke patient rehabilitation conditioning system and method based on a self-learning technology, and relates to the technical field of medical treatment, and the method comprises the steps: collecting illness state data and motion data of a patient, and rehabilitation feedback data in a rehabilitation process; analyzing the collected illness state data, motion data and rehabilitation feedback data based on a self-adaptive intervention strategy; according to an analysis result of the self-learning processing module, a rehabilitation scheme is generated in combination with a multi-process action control framework; wherein the multi-process action control framework comprises the following three processes: forming a change intention, adopting an activity and maintaining the activity. According to the application, the rehabilitation participation degree of the stroke patient is improved through a digital instant self-adaptive intervention scheme.

Description

Technical Field

[0001] This application relates to the field of medical technology, and more particularly, to a rehabilitation conditioning system and method for stroke patients based on self-learning technology. Background Art

[0002] In China, the incidence of stroke remains high and has become the leading cause of death and disability. Early, continuous, and systematic rehabilitation treatment is required after stroke, and among them, motor rehabilitation is an important way for stroke patients to recover. The participation of stroke patients in motor rehabilitation plays a crucial role in improving their own rehabilitation outcomes.

[0003] However, the participation level of stroke patients in motor rehabilitation is low, which seriously affects the functional recovery, quality of life, and disease prognosis of patients after stroke. The participation of home-based stroke patients in motor rehabilitation is affected by multiple internal and external factors, and there is currently no effective intervention program. Summary of the Invention

[0004] The embodiments of this application provide a rehabilitation conditioning system and method for stroke patients based on self-learning technology to solve the above technical problems.

[0005] This application provides a rehabilitation conditioning system for stroke patients based on self-learning technology, including:

[0006] A data acquisition module for collecting the patient's condition data, movement data, and rehabilitation feedback data during the rehabilitation process;

[0007] A self-learning processing module for analyzing the collected condition data, movement data, and rehabilitation feedback data based on an adaptive intervention strategy;

[0008] A rehabilitation plan generation module for generating a rehabilitation plan according to the analysis results of the self-learning processing module in combination with a multi-process action control framework;

[0009] Among them, the multi-process action control framework includes the following three processes: forming the intention to change, adopting activities, and maintaining activities.

[0010] Further, the adaptive intervention strategy includes decision points, decision rules, customization variables, intervention options, intervention plans, and outcome variables; among them,

[0011] The customization variables include learning progress, exercise amount, physiological indicators, and weather information; the outcome variable is the participation degree of motor rehabilitation.

[0012] Further, the self-learning processing module analyzes the collected condition data, movement data, and rehabilitation feedback data based on an adaptive intervention strategy, and is configured to:

[0013] Extract dynamic rehabilitation features from the disease condition data, the exercise data, and the rehabilitation feedback data;

[0014] Based on the dynamic rehabilitation features, use a deep learning algorithm to establish the decision rule and associate the customized variable with the intervention option;

[0015] Utilize the Monte Carlo tree search strategy to optimize the decision rule, and dynamically adjust the decision rule in combination with the value evaluation result of the outcome variable by the deep Q-network;

[0016] Generate the intervention plan according to the dynamic information of the decision rule and the dynamic information of the customized variable;

[0017] Use the deep Q-network to analyze the short-term and long-term effects of the intervention plan, and through the reinforcement learning mechanism, optimize the intervention plan in real time according to the analysis result.

[0018] Furthermore, the dynamic rehabilitation features include muscle strength change features, balance ability change features, exercise volume change features, and pain degree change features;

[0019] Furthermore, the rehabilitation plan generation module generates an intervention plan according to the analysis result of the self-learning processing module and in combination with a multi-process action control framework, and is configured to:

[0020] Take the optimized decision rule and intervention plan output by the self-learning processing module as inputs, and respectively perform operations of forming the intention to change, adopting activities, and maintaining activities in combination with the three processes of the multi-process action control framework;

[0021] Among them, in the process of forming the intention to change, determine the rehabilitation goal according to the optimized decision rule, and generate motivation incentive information and educational materials; in the process of adopting activities, select intervention measures according to the optimized decision rule and the dynamic information of the customized variable, and ensure that the patient completes the rehabilitation training on time through behavior prompts and real-time adjustment; in the process of maintaining activities, set long-term goals according to the patient's rehabilitation progress, and help the patient form a rehabilitation habit through the reinforcement learning mechanism and the social support mechanism;

[0022] Generate an intervention plan including the above three processes.

[0023] Furthermore, the rehabilitation plan generation module includes:

[0024] A decision point setting unit for setting a specific time period of each day as the decision point according to the patient's rehabilitation progress and status;

[0025] A threshold adjustment unit for setting a variable threshold range for the customized variable and adjusting the variable threshold range according to the rehabilitation feedback data.

[0026] Further, the stroke patient rehabilitation conditioning system based on self-learning technology further includes a user interaction module; wherein, the user interaction module includes:

[0027] A learning course unit for displaying exercise rehabilitation courses and activity tasks corresponding to the rehabilitation plan through a mobile application applet;

[0028] A monitoring and recording unit for recording the patient's condition data, exercise data, and rehabilitation feedback data through the mobile application applet;

[0029] A message pushing unit for pushing personalized reminders and suggestions according to the patient's real-time status and the decision rule through the mobile application applet;

[0030] An effect evaluation unit for regularly evaluating the patient's rehabilitation effect through a preset evaluation scale and feeding back the evaluation results to the self-learning processing module to optimize subsequent intervention plans;

[0031] Wherein, the preset evaluation scale includes any one or more of the following: the Stroke Patient Motor Rehabilitation Participation Scale, the 12-Item Short Form Health Survey, and the Modified Barthel Index.

[0032] Further, the exercise rehabilitation courses include start-up and continuous reflection courses, behavior regulation courses, and habit formation courses; wherein,

[0033] The start-up and continuous reflection courses include guidance in aspects such as instrumental attitude, perceived ability, emotional attitude, and perceived opportunity;

[0034] The behavior regulation courses include behavior regulation through action and coping plans, self-monitoring, and self-regulating alternative activities;

[0035] The habit formation courses include guidance in forming habits and identities.

[0036] This application provides a stroke patient rehabilitation conditioning method based on self-learning technology, including: collecting the patient's condition data, exercise data, and rehabilitation feedback data during the rehabilitation process; analyzing the collected condition data, exercise data, and rehabilitation feedback data based on an adaptive intervention strategy; generating a rehabilitation plan according to the analysis results of the self-learning processing module in combination with a multi-process action control framework; wherein, the multi-process action control framework includes the following three processes: forming the intention to change, adopting activities, and maintaining activities.

[0037] Based on the embodiments provided in this application, medical condition data, movement data, and rehabilitation feedback data during the rehabilitation process of patients are collected; the collected medical condition data, movement data, and rehabilitation feedback data are analyzed based on an adaptive intervention strategy; according to the analysis results of the self-learning processing module, combined with a multi-process action control framework, a rehabilitation plan is generated. Thus, through a digital instant adaptive intervention plan, the rehabilitation participation of stroke patients is improved, thereby improving their functional recovery and quality of life; enabling patients to conveniently receive personalized rehabilitation guidance in a home environment, improving the accessibility and quality of rehabilitation services; contributing to optimizing the rehabilitation service system for stroke patients, especially community care services, and further improving the rational allocation and utilization efficiency of rehabilitation resources, promoting social harmony and the improvement of the national health level; exploring the application potential of digital technology in rehabilitation intervention, and promising to provide new solutions for the rehabilitation services of stroke patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, form a part of this application, and the illustrative embodiments and descriptions thereof are used to explain this application and do not constitute an improper limitation to this application. In the drawings:

[0039] Figure 1 FIG. is a structural diagram of an optional stroke patient rehabilitation conditioning system based on self-learning technology according to an embodiment of this application;

[0040] Figure 2 FIG. is a flowchart of an optional stroke patient rehabilitation conditioning method based on self-learning technology according to an embodiment of this application;

[0041] Figure 3 FIG. is a flowchart of another optional stroke patient rehabilitation conditioning method based on self-learning technology according to an embodiment of this application.

[0042] The realization, functional features, and advantages of the objectives of the present invention will be further described in conjunction with the embodiments and with reference to the drawings. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0043] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the scope of protection of the present invention.

[0044] Stroke is the leading cause of death and disability in China, and its prevention and control are the current focus. Early, continuous, and systematic rehabilitation treatment is of great significance for the functional recovery and quality of life improvement of stroke patients. Among them, exercise rehabilitation is an important way for stroke recovery.

[0045] The previous conceptual analysis results of the research group on stroke showed that the rehabilitation participation of patients is a dynamic and evolving process and state, which requires patients to maintain positive emotions, continuous commitment and actions, and actively communicate and cooperate with rehabilitation service providers to ensure their continuous adaptation to the rehabilitation plan and achieve rehabilitation goals. However, the level of exercise rehabilitation participation of stroke patients is far lower than the expected level, seriously affecting the functional recovery, quality of life and disease prognosis of patients after stroke. The exercise rehabilitation participation of home-based stroke patients is affected by multiple internal and external factors, and there is currently no effective intervention program.

[0046] There is a shortage and uneven distribution of rehabilitation medical resources in China. Digital health management provides new ideas and perspectives for the exercise rehabilitation intervention of stroke patients. Developing an exercise rehabilitation participation management program and intervention plan for stroke patients has good exploration value, but there is currently a lack of relevant application research.

[0047] In view of the particularity and importance of disease management in the stroke population, it is necessary to deeply understand: how to provide nursing interventions through digital tools to improve the level of patients' exercise rehabilitation participation and thus improve health outcomes.

[0048] Therefore, this application is guided by M-PAC and intends to construct an immediate adaptive intervention plan for the exercise rehabilitation participation of home-based stroke patients. Relying on a mobile application mini-program, a randomized controlled trial is carried out to verify the effect of the intervention plan. The research results will help to improve the exercise rehabilitation participation of home-based stroke patients, thereby improving the rehabilitation effect and quality of life of patients, and providing a useful exploration for improving the rehabilitation system of stroke patients in China and enhancing the accessibility and quality of rehabilitation services.

[0049] Among them, just-in-time adaptive interventions (JITAIs) aim to provide timely support and dynamically meet the needs of users by adapting to the individual's internal state and environmental dynamics, including five key components: decision points, decision rules, customization variables, intervention plans, and outcome variables. Just-in-time adaptive interventions have great potential in promoting healthy behavior change and have been effectively applied in promoting physical activity, substance abuse, etc.

[0050] To promote patients to bridge the intention-behavior gap in physical activity, a multi-process action control (M-PAC) framework can be constructed. M-PAC is a hierarchical behavior change method that involves three interrelated but sequential processes: the reflection process, the regulation process, and the reflex process. The physical activity intervention based on this framework is an effective and cutting-edge technology. At the same time, existing research has proposed combining M-PAC and immediate adaptive intervention for clinical application, but the above technologies have not been effectively explored in China.

[0051] Optionally, as Figure 1 shown, this application provides a rehabilitation conditioning system for stroke patients based on self-learning technology, including:

[0052] A data collection module 101, which is used to collect the patient's condition data, movement data, and rehabilitation feedback data during the rehabilitation process; among them, the movement data is dynamically collected through a sports bracelet, and the real-time feedback data is obtained through the interaction between the patient and the system;

[0053] A self-learning processing module 102, which is used to analyze the collected condition data, movement data, and rehabilitation feedback data based on an adaptive intervention strategy;

[0054] A rehabilitation plan generation module 103, which is used to generate a rehabilitation plan according to the analysis results of the self-learning processing module, in combination with the multi-process action control framework;

[0055] Among them, the M-PAC framework (Multi-Process Action Control Framework) is a theoretical framework for understanding and explaining human behavior motivation, decision-making, and action control. It is mainly used to explain how individuals regulate and control their behavior through multiple psychological processes when facing complex situations; the multi-process action control framework includes the following three processes: forming the intention to change, adopting activities, and maintaining activities.

[0056] Based on the embodiments provided in this application, the condition data, movement data, and rehabilitation feedback data of the patient are collected; the collected condition data, movement data, and rehabilitation feedback data are analyzed based on an adaptive intervention strategy; according to the analysis results of the self-learning processing module, combined with a multi-process action control framework, a rehabilitation plan is generated. Thus, through a digital instant adaptive intervention plan, the rehabilitation participation of stroke patients is improved, thereby improving their functional recovery and quality of life; enabling patients to conveniently receive personalized rehabilitation guidance in a home environment, improving the accessibility and quality of rehabilitation services; helping to optimize the rehabilitation service system for stroke patients, especially community nursing services, and further improving the rational allocation and utilization efficiency of rehabilitation resources, promoting social harmony and the improvement of the national health level; exploring the application potential of digital technology in rehabilitation intervention, and promising to provide new solutions for the rehabilitation services of stroke patients.

[0057] Further, the adaptive intervention strategy includes decision points, decision rules, customization variables, intervention options, intervention plans, and outcome variables; among them,

[0058] The customization variables include learning progress, exercise volume, physiological indicators, and weather information; the outcome variable is the participation in exercise rehabilitation.

[0059] Further, as Figure 2 shown, the self-learning processing module analyzes the collected condition data, movement data, and rehabilitation feedback data based on the adaptive intervention strategy, and is configured as:

[0060] S201, extract dynamic rehabilitation features from the condition data, movement data, and rehabilitation feedback data;

[0061] For example, by calculating the muscle strength difference between adjacent time points, the muscle strength change rate can be obtained, which is used to evaluate the patient's rehabilitation progress;

[0062] S202, based on the dynamic rehabilitation features, use a deep learning algorithm to establish decision rules, and associate the customization variables with the intervention options;

[0063] For example, according to the patient's muscle strength change rate and exercise volume change rate, decide whether to adjust the exercise intensity or increase the frequency of rehabilitation training;

[0064] S203, use the Monte Carlo tree search strategy to optimize the decision rules, and dynamically adjust the decision rules in combination with the value evaluation results of the outcome variables by the deep Q network;

[0065] Among them, the full name of the deep Q network is Deep Q-Network. It is an algorithm that combines the deep neural network in deep learning and the Q-Learning algorithm in reinforcement learning.

[0066] S204. Generate an intervention plan based on the dynamic information of the decision rule and the dynamic information of the customized variables.

[0067] S205. Analyze the short-term and long-term effects of the intervention plan using a deep Q-network, and optimize the intervention plan in real time according to the analysis results through a reinforcement learning mechanism.

[0068] Based on the embodiments provided in this application, the self-learning processing module extracts dynamic rehabilitation features from the patient's condition data, movement data, and rehabilitation feedback data. These features include muscle strength change features, balance ability change features, exercise volume change features, and pain degree change features, etc. The extraction of these features enables the system to understand the patient's rehabilitation progress and status in real time. Based on the extracted dynamic rehabilitation features, decision rules are established using deep learning algorithms. These rules associate customized variables, such as learning progress, exercise volume, physiological indicators, and weather information, with intervention options. This enables the system to dynamically adjust intervention measures according to the specific situation of the patient. The Monte Carlo tree search strategy is used to optimize the decision rules. By simulating multiple random paths, the potential effects of different decision rules are evaluated to ensure the selection of the optimal intervention strategy. According to the optimized decision rules and the dynamic information of the customized variables, personalized intervention plans are generated. These plans can be adjusted in real time to adapt to the changing needs of the patient. The short-term and long-term effects of the intervention plan are analyzed using a deep Q-network, and the intervention plan is optimized in real time through a reinforcement learning mechanism. This ensures the continuous improvement and effectiveness of the intervention plan.

[0069] Thus, it is achieved that: through the extraction of dynamic rehabilitation features and the optimization of decision rules, accurate dynamic intervention is provided, significantly improving the patient's participation and effect in motor rehabilitation. This accurate intervention can adjust the rehabilitation plan according to the patient's real-time state, ensuring the efficiency and pertinence of the rehabilitation process. Based on the patient's subjective feedback and objective data, the system can provide personalized rehabilitation guidance, enhancing the patient's sense of self-efficacy and motivation for continuous participation. This personalized guidance can help the patient better understand and execute the rehabilitation plan, improving the compliance of rehabilitation. By optimizing the intervention plan in real time, the system can improve the rehabilitation efficiency, shorten the rehabilitation cycle, and reduce medical costs. This real-time optimization can ensure that the rehabilitation plan is always in the best state, reducing ineffective intervention and resource waste during the rehabilitation process.

[0070] Furthermore, the dynamic rehabilitation features include muscle strength change features, balance ability change features, exercise volume change features, and pain degree change features;

[0071] Among them, the Monte Carlo tree search strategy evaluates the potential effects of different decision rules by simulating multiple random paths. The specific formula is as follows:

[0072]

[0073] Among them, V node (s) represents the value evaluation of the node in state s, that is, the potential value of taking different intervention options in this state; R(s, a i ) represents the reward value of taking intervention option a i in state s. For example, the degree of muscle strength improvement and pain reduction of the patient, etc., can be initially estimated by the deep Q-network; w i is the weight coefficient, used to adjust the reward value weights of different intervention options; N represents the total number of simulations, that is, the number of simulations of different intervention paths in the Monte Carlo tree search strategy; i is the index of the simulation times; N(s) represents the number of times state s is visited, used to measure the attention of this state during the simulation process; c represents the exploration parameter, used to balance exploration and exploitation, and can be adjusted according to actual needs. For example, at the initial stage of rehabilitation, the value of c can be appropriately increased to increase exploration; M represents the number of customized variables; j is the index of the customized variable; δ j is the change rate of the jth customized variable, used to reflect the current rehabilitation dynamics of the patient, such as the muscle strength change rate, exercise volume change rate, etc.; d j is the importance coefficient of the jth customized variable, set according to the characteristics of stroke rehabilitation and clinical experience. For example, if the importance of muscle strength for rehabilitation may be higher than that of exercise volume, then d j is correspondingly larger.

[0074] Based on the embodiments provided in this application, the Monte Carlo tree search strategy evaluates the potential effects of different decision rules in actual rehabilitation by simulating multiple intervention paths. This kind of simulation can balance the relationship between exploring new strategies and exploiting known effective strategies, and avoid falling into local optima. That is to say, this kind of simulation can help the system predict the possible results of intervention plans in advance, so as to select the optimal intervention strategy. For example, if it is found in the simulation that a certain intervention measure has a significant effect on the muscle strength recovery of the patient at a specific stage, the system can preferentially select this measure.

[0075] Through the extraction of dynamic rehabilitation features and the optimization of the Monte Carlo tree search strategy, the system can dynamically adjust the intervention plan according to the actual rehabilitation situation of the patient. This adaptability enables the system to still provide effective rehabilitation guidance when facing different individual differences of patients. The Monte Carlo tree search strategy evaluates the effects of intervention plans through multiple simulations, avoiding the fluctuations in rehabilitation effects caused by single decisions. The system can maintain stable intervention effects during the rehabilitation process, which helps patients build rehabilitation confidence and improve the persistence of rehabilitation.

[0076] In the complex situation of stroke rehabilitation conditioning, the Monte Carlo tree search strategy can comprehensively consider various factors, such as the patient's physiological indicators, rehabilitation progress, etc., and make intelligent decisions. This intelligent decision-making ability enables the system to maintain efficient and accurate intervention in the face of complex and changing rehabilitation situations. Specifically, the introduction of the exploration parameter c enables the system to balance between exploring new strategies and exploiting known effective strategies, avoiding falling into local optima. This balance ensures that the system continuously explores new intervention methods during the rehabilitation process while fully utilizing known effective strategies, improving the comprehensiveness and adaptability of rehabilitation. Through the extraction of dynamic rehabilitation features and the optimization of the Monte Carlo tree search strategy, the system can dynamically adjust the intervention plan to adapt to the changing needs of patients, improving the adaptability and flexibility of the intervention. This adaptability ensures that the system always provides the most appropriate intervention measures during the patient's rehabilitation process, improving the sustainability and effectiveness of rehabilitation.

[0077] Furthermore, the final analysis results of the short-term and long-term effects of the intervention plan are determined based on the following formula:

[0078]

[0079] where Q intervention (s, pl) represents the evaluation value of the final intervention effect of adopting the intervention plan pl in state s, which is used to guide the optimization of the intervention plan; Q base (s, pl) is the basic Q value obtained by the deep Q network through learning historical data, indicating the effect of adopting the intervention plan pl in state s under normal circumstances; σ is an adjustment coefficient used to control the influence degree of additional evaluation factors on the Q value, which can be adjusted according to the rehabilitation stage and actual situation; T represents the number of time periods for short-term effect evaluation, such as daily, weekly, etc.; k is the index of the time period for short-term effect evaluation; Δ k is the change amount of the patient's rehabilitation indicators within the k-th time period, such as the improvement amplitude of muscle strength, joint range of motion, etc., which is used to evaluate the short-term effect of the intervention measure; f k is the weight coefficient of the k-th time period, which can be set according to the importance degree of the time period. For example, the weight in the initial stage of rehabilitation may be higher; U represents the number of factors for long-term effect evaluation, such as post-rehabilitation exercise habits, quality of life, etc.; l is the index of the factor for long-term effect evaluation; θ l is the change rate of the l-th long-term effect evaluation factor, such as the change rate of exercise frequency after rehabilitation, the change rate of quality of life score, etc.; g l is the importance coefficient of the l-th long-term effect evaluation factor; it is set according to the long-term goals of stroke rehabilitation and the patient's needs. For example, if exercise habits are more important for long-term rehabilitation than quality of life, then g l is correspondingly larger.

[0080] During the process of stroke rehabilitation conditioning, short-term effect evaluation can timely feedback the actual effects of the intervention plan in the short term. For example, by evaluating indicators such as muscle strength improvement and balance ability improvement of patients at a certain stage, the system can quickly judge whether the intervention plan is effective. This helps the system timely adjust the intervention measures and avoid the continuous implementation of ineffective interventions. Long-term effect evaluation focuses on the impact of the intervention plan on the overall rehabilitation process of patients. By analyzing indicators such as functional recovery and quality of life improvement of patients in the long term, the system can comprehensively evaluate the long-term value of the intervention plan. This helps the system optimize the intervention plan to ensure its continuous effectiveness in the long-term rehabilitation process.

[0081] Based on the embodiments provided in this application, through the comprehensive evaluation of short-term and long-term effects, the system can comprehensively understand the effects of the intervention plan in different time dimensions. This comprehensive evaluation enables the system to optimize the intervention plan from both short-term and long-term perspectives, improving the comprehensiveness and sustainability of rehabilitation. Long-term effect evaluation can ensure that the intervention plan is not only effective in the short term but also can continuously improve the rehabilitation status of patients in the long term. By continuously optimizing the intervention plan, the system improves the sustainability of rehabilitation, which helps patients achieve long-term rehabilitation goals.

[0082] Among them, the deep Q-network can provide intelligent decision-making support for the adjustment of stroke rehabilitation intervention plans based on a large amount of historical data and real-time feedback. It helps the system select the optimal intervention strategy by learning the effects of different intervention measures at different stages. For example, the deep Q-network can analyze the movement data and rehabilitation feedback of patients at different rehabilitation stages and predict which intervention measure is most likely to improve the patient's movement rehabilitation participation and effect. The deep Q-network can dynamically evaluate the short-term and long-term effects of the intervention plan. It can predict the potential effects of different intervention measures based on the patient's real-time data and historical data, thus helping the system timely adjust the intervention plan. For example, the deep Q-network can analyze the rehabilitation data of patients at a certain stage and predict the short-term and long-term effects of continuing the current intervention measure, so as to decide whether to adjust the intervention plan.

[0083] The deep Q-network can provide personalized rehabilitation guidance according to the individual differences of patients. It tailors a rehabilitation plan for each patient by analyzing data such as the patient's physiological indicators and rehabilitation progress, improving the pertinence and effectiveness of rehabilitation. For example, the deep Q-network can adjust the intervention plan according to the dynamic rehabilitation characteristics of patients such as muscle strength changes and balance ability changes to ensure that each patient can obtain the most suitable rehabilitation guidance for themselves.

[0084] Specifically, during the process of stroke rehabilitation, assume that the rehabilitation state of the patient at a certain stage is s t , and the intervention measure selected by the system is a t . By executing a t , the patient enters a new state st+1 and obtain a reward r t+1 (e.g., the reward for muscle strength improvement). The deep Q-network adjusts the intervention plan by updating the Q-value. If an intervention measure can bring a high reward in multiple stages, the deep Q-network will preferentially select this measure to ensure the optimality of the intervention plan.

[0085] Furthermore, the rehabilitation plan generation module generates an intervention plan according to the analysis result of the self-learning processing module and in combination with the multi-process action control framework, and is configured to:

[0086] Take the optimized decision rule and intervention plan output by the self-learning processing module as inputs, and respectively perform operations of forming the intention to change, adopting activities, and maintaining activities in combination with the three processes of the multi-process action control framework;

[0087] Among them, in the process of forming the intention to change, determine the rehabilitation goal according to the optimized decision rule, and generate motivation incentive information and educational materials; in the process of adopting activities, select intervention measures according to the optimized decision rule and the dynamic information of the customized variables, and ensure that the patient completes the rehabilitation training on time through behavior prompts and real-time adjustment; in the process of maintaining activities, set long-term goals according to the patient's rehabilitation progress, and help the patient form a rehabilitation habit through the reinforcement learning mechanism and the social support mechanism;

[0088] Generate an intervention plan including the above three processes.

[0089] Among them, the intervention measure is selected based on the following formula:

[0090] I(t) = argmax m∈M [U(m,t) + V(m,t)]

[0091] where I(t) is the intervention measure selected at time t; M is the set of intervention measures; U(m,t) is the utility value of taking the intervention measure m at time t; V(m,t) is the value evaluation value of taking the intervention measure m at time t determined according to the deep Q-network;

[0092] Furthermore, the rehabilitation plan generation module includes:

[0093] A decision point setting unit, configured to set a specific time period of each day as a decision point according to the patient's rehabilitation progress and status;

[0094] A threshold adjustment unit, configured to set a variable threshold range for the customized variable and adjust the variable threshold range according to the rehabilitation feedback data.

[0095] The self-learning processing module further includes: establishing a decision rule based on a machine learning algorithm to adapt to the patient's individual internal state and environmental dynamics, and providing timely support and intervention suggestions.

[0096] The self-learning processing module further includes: a variable customization unit: setting customized variables as learning progress, exercise amount, physiological indicators, weather information, etc., to comprehensively reflect the patient's state and environmental factors; a knowledge base management unit: storing knowledge and experience related to stroke rehabilitation, providing reference and basis for the generation of intervention plans.

[0097] Furthermore, the stroke patient rehabilitation conditioning system based on self-learning technology further includes a user interaction module; wherein, the user interaction module includes:

[0098] A learning course unit, used to display exercise rehabilitation courses and activity tasks corresponding to the rehabilitation plan through a mobile application mini-program;

[0099] A monitoring and recording unit, used to record the patient's condition data, exercise data, and rehabilitation feedback data through a mobile application mini-program;

[0100] A message push unit, used to push personalized reminders and suggestions according to the patient's real-time state and decision rules through a mobile application mini-program;

[0101] An effect evaluation unit, used to regularly evaluate the patient's rehabilitation effect through a preset evaluation scale, and feedback the evaluation results to the self-learning processing module to optimize subsequent intervention plans;

[0102] Among them, the preset evaluation scale includes any one or more of the following: the Stroke Patient Motor Rehabilitation Participation Scale, the 12-Item Short Form Health Survey, and the Modified Barthel Index.

[0103] Furthermore, the exercise rehabilitation courses include start-up and continuous reflection courses, behavior regulation courses, and habit formation courses; wherein,

[0104] The start-up and continuous reflection courses include guidance in aspects such as instrumental attitude, perceived ability, emotional attitude, and perceived opportunity;

[0105] The behavior regulation courses include behavior regulation through action and coping plans, self-monitoring, and self-regulation alternative activities;

[0106] The habit formation courses include guidance in forming habits and identities.

[0107] Optionally, as Figure 3 shown, this application provides a stroke patient rehabilitation conditioning method based on self-learning technology, including:

[0108] S301, collecting the patient's condition data, exercise data, and rehabilitation feedback data during the rehabilitation process;

[0109] S302. Analyze the collected condition data, exercise data, and rehabilitation feedback data based on the adaptive intervention strategy;

[0110] S303. Generate a rehabilitation plan according to the analysis results of the self-learning processing module and in combination with the multi-process action control framework;

[0111] Among them, the multi-process action control framework includes the following three processes: forming the intention to change, adopting activities, and maintaining activities.

[0112] It should be noted that the embodiments of the present application are based on the previously clarified concept of patient participation in rehabilitation and the home-based exercise rehabilitation participation scale for stroke patients, and under the guidance of the multi-process action control (M-PAC) framework, an intervention plan for home-based stroke patient exercise rehabilitation participation is constructed. A sports bracelet is used to dynamically collect and feedback patients' exercise data. By developing a sports rehabilitation management mini-program and combining the just-in-time adaptive interventions (JITAIs) mode, precise dynamic intervention is achieved, and a randomized controlled trial is carried out to verify the effect of the intervention plan. The research results will help improve the level of home-based stroke patient exercise rehabilitation participation and physical activity volume, provide effective guidance and support for home-based stroke patient exercise rehabilitation, and are of great significance for improving the stroke patient rehabilitation system in China and enhancing the accessibility and quality of rehabilitation services.

[0113] Specifically, in some embodiments of the present application, the mini-program is developed with a patient side and a management side. Through literature review and expert consultation, a detailed intervention plan is formulated, including 10 courses, each course lasting 20 - 30 minutes, covering multiple aspects such as starting reflection, behavior regulation, self-monitoring, and habit formation. Patients promote the transformation of awareness and behavior through course learning. Provide patients with exercise prescriptions, combine the just-in-time adaptive interventions mode, and instantaneously adjust the dose of exercise intervention according to patients' subjective feedback and objective data to achieve visual feedback of data, provide personalized rehabilitation guidance, and thus achieve precise dynamic intervention. In order to verify the effect of the intervention plan, the method of randomized controlled trial is used for verification. It is expected that through this innovative method, the exercise rehabilitation participation of patients can be significantly improved, their rehabilitation outcomes can be improved, and effective guidance and support can be provided for home-based rehabilitation of stroke patients. The research results not only have important theoretical and practical values, but also have broad application prospects and social significance, and are expected to provide a new solution for the rehabilitation service system of stroke patients in China, promote the digital transformation of rehabilitation services, and improve the rational allocation and utilization efficiency of rehabilitation resources.

[0114] The embodiments of this application are based on the M-PAC framework, providing a comprehensive theoretical perspective for the research. A set of immediate adaptive intervention programs are designed based on the literature. This program is implemented through a mini-program and can adjust the rehabilitation plan and support strategy immediately according to the patient's subjective feedback and needs and objective data collection. The intervention program covers multiple aspects such as cognitive education, emotional support, and behavior motivation, aiming to enhance the patient's awareness of the importance of rehabilitation, stimulate positive emotions, enhance self-efficacy, and promote the patient's continuous participation in exercise rehabilitation.

[0115] The embodiments of this application are based on the M-PAC framework and apply it to the intervention of stroke patients' participation in exercise rehabilitation, which has theoretical innovation; this study constructs a mini-program based on the immediate adaptive intervention model, which can provide reference for the research and practice of stroke health management.

[0116] Through targeted intervention measures, it is expected to improve the participation of stroke patients in exercise rehabilitation in the home environment. It is expected that the intervention program can motivate patients to increase physical activity, thus having a positive impact on the rehabilitation process. The research results are expected to provide more effective home exercise rehabilitation guidance and support for stroke patients, thereby improving the patient's functional recovery, quality of life, and prognosis of the disease.

[0117] It should be noted that in this application, the embodiments implemented on the side of the stroke patient rehabilitation conditioning system based on self-learning technology can be referred to each other with the embodiments implemented on the side of the stroke patient rehabilitation conditioning method based on self-learning technology, and this application will not elaborate on them one by one.

[0118] The above are only the preferred embodiments of the present invention, and do not limit the patent scope of the present invention accordingly. Any equivalent structural or equivalent process transformation made by using the content of the specification and drawings of the present invention, or directly or indirectly applied in other related technical fields, shall be equally included in the patent protection scope of the present invention.

Claims

1. A rehabilitation conditioning system for stroke patients based on self-learning technology, characterized in that, Including: A data acquisition module, configured to acquire the patient's condition data, movement data, and rehabilitation feedback data during the rehabilitation process; A self-learning processing module, configured to analyze the acquired condition data, movement data, and rehabilitation feedback data based on an adaptive intervention strategy; A rehabilitation plan generation module, configured to generate a rehabilitation plan according to the analysis result of the self-learning processing module and in combination with a multi-process action control framework; Wherein, the multi-process action control framework includes the following three processes: forming an intention to change, adopting an activity, and maintaining an activity.

2. The rehabilitation conditioning system for stroke patients based on self-learning technology according to claim 1, wherein The adaptive intervention strategy includes decision points, decision rules, customization variables, intervention options, intervention plans, and outcome variables; Wherein, the customization variables include learning progress, exercise amount, physiological indicators, and weather information; the outcome variable is the participation degree of exercise rehabilitation.

3. The rehabilitation conditioning system for stroke patients based on self-learning technology according to claim 2, wherein The self-learning processing module analyzes the acquired condition data, movement data, and rehabilitation feedback data based on an adaptive intervention strategy, and is configured to: Extract dynamic rehabilitation features from the condition data, movement data, and rehabilitation feedback data; Based on the dynamic rehabilitation features, establish the decision rules using a deep learning algorithm, and associate the customization variables with the intervention options; Optimize the decision rules using a Monte Carlo tree search strategy, and dynamically adjust the decision rules in combination with the value evaluation result of the outcome variable by a deep Q-network; Generate the intervention plan according to the dynamic information of the decision rules and the dynamic information of the customization variables; Analyze the short-term and long-term effects of the intervention plan using the deep Q-network, and optimize the intervention plan in real time according to the analysis result through a reinforcement learning mechanism.

4. The rehabilitation conditioning system for stroke patients based on self-learning technology according to claim 3, characterized in that, The dynamic rehabilitation features include muscle strength change features, balance ability change features, exercise amount change features, and pain degree change features; Wherein, the Monte Carlo tree search strategy evaluates the potential effects of different decision rules by simulating multiple random paths.

5. The rehabilitation conditioning system for stroke patients based on self-learning technology according to claim 3, characterized in that, Determine the final analysis result of the short-term and long-term effects of the intervention plan.

6. The rehabilitation conditioning system for stroke patients based on self-learning technology according to claim 3, wherein The rehabilitation plan generation module generates an intervention plan according to the analysis result of the self-learning processing module and in combination with a multi-process action control framework, and is configured to: Take the optimized decision rules and intervention plan output by the self-learning processing module as inputs, and respectively perform operations of forming an intention to change, adopting an activity, and maintaining an activity in combination with the three processes of the multi-process action control framework; Wherein, in the process of forming an intention to change, determine the rehabilitation goal according to the optimized decision rules, and generate motivation incentive information and educational materials; in the process of adopting an activity, select intervention measures according to the optimized decision rules and the dynamic information of the customization variables, and ensure that the patient completes the rehabilitation training on time through behavior prompts and real-time adjustment; in the process of maintaining an activity, set long-term goals according to the patient's rehabilitation progress, and help the patient form a rehabilitation habit through a reinforcement learning mechanism and a social support mechanism; Generate an intervention plan including the above three processes.

7. The rehabilitation conditioning system for stroke patients based on self-learning technology according to claim 2, characterized in that, The rehabilitation plan generation module includes: A decision point setting unit for setting a specific time period of each day as the decision point according to the patient's rehabilitation progress and status; A threshold adjustment unit for setting a variable threshold range for the customization variable and adjusting the variable threshold range according to the rehabilitation feedback data.

8. The rehabilitation conditioning system for stroke patients based on self-learning technology according to claim 2, characterized in that, The stroke patient rehabilitation conditioning system based on self-learning technology further includes a user interaction module; wherein, the user interaction module includes: A learning course unit for displaying exercise rehabilitation courses and activity tasks corresponding to the rehabilitation plan through a mobile application applet; A monitoring and recording unit for recording the patient's condition data, exercise data and rehabilitation feedback data through the mobile application applet; A message pushing unit for pushing personalized reminders and suggestions according to the patient's real-time status and the decision rule through the mobile application applet; An effect evaluation unit for regularly evaluating the patient's rehabilitation effect through a preset evaluation scale and feeding back the evaluation results to the self-learning processing module to optimize subsequent intervention plans; Wherein, the preset evaluation scale includes any one or more of the following: Stroke Patient Motor Rehabilitation Participation Scale, 12-Item Short Form Health Survey, and Modified Barthel Index.

9. The rehabilitation conditioning system for stroke patients based on self-learning technology according to claim 8, characterized in that, The exercise rehabilitation courses include start-up and continuous reflection courses, behavior regulation courses, and habit formation courses; Wherein, the start-up and continuous reflection courses include guidance in aspects such as instrumental attitude, perceived ability, emotional attitude, and perceived opportunity; The behavior regulation courses include behavior regulation through actions and coping plans, self-monitoring, and self-regulating alternative activities; The habit formation courses include guidance in forming habits and identities.

10. A rehabilitation conditioning method for stroke patients based on self-learning technology, characterized in that, Including: Collecting the patient's condition data, exercise data, and rehabilitation feedback data during the rehabilitation process; Analyzing the collected condition data, exercise data, and rehabilitation feedback data based on an adaptive intervention strategy; Generating a rehabilitation plan according to the analysis results of the self-learning processing module and in combination with a multi-process action control framework; Wherein, the multi-process action control framework includes the following three processes: forming the intention to change, adopting activities, and maintaining activities.

Images