Biofeedback-driven personalized rehabilitation optimization method

Through the integration of multimodal biofeedback data and self-report data, a comprehensive health status assessment model is constructed, personalized rehabilitation goals are set, and intelligent rehabilitation plans are generated, which solves the problems of insufficient personalization and unclear goals of the existing biofeedback system. An efficient and personalized rehabilitation training plan has been achieved, which has significantly improved the rehabilitation effect.

CN119028582AInactive Publication Date: 2024-11-26SANYA HOSPITAL OF TRADITIONAL CHINESE MEDICINE

Patent Information

Application Number
CN202411085914.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-08
Publication Date
2024-11-26
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The existing biofeedback system only monitors a single physiological indicator, and cannot achieve comprehensive personalized rehabilitation optimization. The lack of comprehensive assessment of the patient's comprehensive health status makes it difficult for the rehabilitation plan to accurately reflect the patient's overall health status, and it is difficult to adjust dynamically. Not clear, the rehabilitation effect is difficult to guarantee.

Method used

By collecting multimodal biofeedback data and self-report data, using a multi-dimensional data fusion algorithm to build a comprehensive health status assessment model, setting personalized rehabilitation goals, and generating intelligent rehabilitation programs, including exercise training, physical therapy, psychological counseling and nutritional advice. Optimize rehabilitation programs using a hybrid model of transfer learning and self-organizing mapping.

Benefits of technology

The scientificity and accuracy of the rehabilitation plan are improved, and highly personalized rehabilitation training plans can be generated according to the specific needs of each patient, ensuring the continuity and efficiency of the rehabilitation process, dynamically adjusting the rehabilitation goals, and significantly improving the rehabilitation effect.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119028582A_ABST
    Figure CN119028582A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of rehabilitation methods, in particular to a biofeedback-driven personalized rehabilitation optimization method, which comprises the following steps of: acquiring multi-modal biofeedback data of a patient, and acquiring self-report data of the patient through user terminal equipment; integrating data from different sources by using a multi-dimensional data fusion algorithm, constructing a comprehensive health state evaluation model, and identifying key health indexes; personalized rehabilitation targets are set according to the recognized key health indexes, and the rehabilitation targets comprise short-term rehabilitation targets and long-term rehabilitation targets; based on the short-term and long-term rehabilitation targets, using an intelligent algorithm to generate a personalized rehabilitation training scheme; according to the method, clear short-term and long-term rehabilitation targets are set according to the identified key health indexes, the short-term targets pay attention to rapid improvement of the current health state of the patient, and the target-oriented rehabilitation method ensures the continuity and high efficiency of the rehabilitation process.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of rehabilitation methods, and in particular to a biofeedback-driven personalized rehabilitation optimization method. Background Art

[0002] Rehabilitation therapy plays an important role in the medical field, aiming to help patients recover their health and improve their quality of life. Traditional rehabilitation methods rely on standardized treatment plans and lack individual considerations, resulting in uneven rehabilitation effects. In recent years, biofeedback technology has been gradually applied to rehabilitation therapy, which improves the effect of rehabilitation training by real-time monitoring of patients' physiological indicators and providing feedback. However, existing biofeedback systems often only focus on a single indicator and cannot achieve comprehensive personalized rehabilitation optimization. The following technical problems exist:

[0003] Existing biofeedback systems usually only monitor single physiological indicators such as heart rate and electromyography, and lack a comprehensive assessment of the patient's overall health status. The limitations of this single data source make it difficult for rehabilitation programs to accurately reflect the patient's overall health status and cannot provide comprehensive treatment for multiple health problems.

[0004] Lack of personalization and difficulty in dynamic adjustment: Traditional rehabilitation programs lack personalized considerations and mostly use fixed training intensity and content. They cannot be dynamically adjusted based on real-time feedback from patients. Inflexible programs are difficult to adapt to the rehabilitation needs of different patients and may result in excessive or insufficient training intensity, affecting the rehabilitation effect.

[0005] Unclear goals make it difficult to guarantee rehabilitation results: Many rehabilitation programs lack clear short-term and long-term rehabilitation goals, resulting in a lack of direction and phased evaluation in the rehabilitation process. Without clear rehabilitation goals, rehabilitation training can hardly focus on solving key health problems, and rehabilitation results are difficult to guarantee. Summary of the invention

[0006] Based on the above objectives, the present invention provides a biofeedback-driven personalized rehabilitation optimization method.

[0007] The biofeedback-driven personalized rehabilitation optimization method includes the following steps:

[0008] S1, data collection: collect multimodal biofeedback data from patients, including heart rate variability, galvanic skin response, electromyographic signals, breathing patterns, and brain waves, and collect self-reported data from patients through user terminal devices, including pain level, emotional state, and fatigue;

[0009] S2, multi-dimensional data fusion and preliminary analysis: pre-process the collected multi-modal biofeedback data and self-report data, use multi-dimensional data fusion algorithm to integrate data from different sources, build a comprehensive health status assessment model, evaluate the overall health status of patients based on the comprehensive health status assessment model, and identify key health indicators;

[0010] S3, personalized rehabilitation goal setting: according to the identified key health indicators, set personalized rehabilitation goals, including short-term and long-term rehabilitation goals;

[0011] S4, intelligent rehabilitation program formulation: Based on short-term and long-term rehabilitation goals, use intelligent algorithms to generate personalized rehabilitation training programs, which include sports training, physical therapy, psychological counseling, and nutritional advice.

[0012] Optionally, the S1 specifically includes:

[0013] S11, heart rate variability data collection: use a wearable heart rate monitoring device (heart rate belt or smart watch) to monitor the patient's heart rate data in real time, record and analyze the time interval of the heart rate, i.e., the RR interval, through the data acquisition module, and calculate the heart rate variability (HRV) index;

[0014] S12, skin electrical response data collection: Use skin potential sensors attached to the patient's palm or fingertips to measure the changes in skin conductivity in real time and record skin electrical response data to reflect the patient's emotions and stress status;

[0015] S13, electromyographic signal data acquisition: Use surface electromyography (sEMG) sensors attached to the patient's target muscle groups to record the electrical activity signals of the muscles in real time, monitor muscle contraction and relaxation, and evaluate muscle function and fatigue status;

[0016] S14, breathing pattern data acquisition: using a breathing belt to monitor the patient's breathing rate and depth, and recording breathing pattern data, including breathing rate, through a data acquisition module;

[0017] S15, EEG data collection: Use a portable electroencephalogram (EEG) device to monitor the patient's brain wave activity, record brain wave signals in real time, and analyze EEG activities in different frequency bands;

[0018] S16, self-report data collection: A user interface is provided through the user terminal device (smartphone or tablet computer), through which the patient enters the self-report data, and regular reminders are set to remind the patient to fill in the pain score, emotional state and fatigue data.

[0019] Optionally, the S2 specifically includes:

[0020] S21, data preprocessing: filtering technology was used to remove noise from heart rate variability, galvanic skin response, electromyographic signals, breathing patterns, and brain wave data. Each biofeedback data was standardized so that different types of data had the same dimension and range. The z-score standardization method was used to convert the data into a standard normal distribution with a mean of 0 and a standard deviation of 1.

[0021] S22, using feature selection and feature extraction techniques, extract features from data from different sources, construct feature vectors, and use independent component analysis dimensionality reduction algorithm to reduce data dimensions and reduce redundant information. Let the original data matrix be X∈R m×n , where m is the number of samples and n is the number of features. Independent component analysis is used to reduce the dimension of the standardized data and extract independent components. The standardized data is expressed as: Z = [z1, z2, ..., z n ], using independent component analysis Z = AS, where A is the mixing matrix and S is the independent component matrix, and independent components are extracted by solving S;

[0022] S23, based on a multimodal fusion neural network, the standardized biofeedback data and self-report data are integrated to construct a comprehensive health status assessment model to reflect the overall health status of the patient. The multimodal fusion neural network takes different types of biofeedback data and self-report data as input, and the input vector is I = [i1, i2, ..., i k ], where k is the number of data types. The network structure of the multimodal fusion neural network includes an input layer, multiple hidden layers, and an output layer. The input layer receives the fused feature vector, the hidden layer performs nonlinear transformation, and the output layer generates a health status score. ;

[0023] S24, health status assessment: input the real-time collected biofeedback data and self-report data into the comprehensive health status assessment model, and based on the model output results, evaluate the patient's overall health status and generate a health status score;

[0024] S25, Identify key health indicators: By analyzing network weights and feature contribution analysis, identify key health indicators that affect patients' health. Key health indicators include heart rate variability, electromyographic signal strength, galvanic skin response, electromyographic signal, respiratory rate, brain waves, self-reported pain level, self-reported emotional state, and self-reported fatigue.

[0025] Optionally, the comprehensive health status score (i.e., the comprehensive health status assessment model expression) is expressed as: Where W1, W2, W3 are weight matrices, b1, b2, b3 are bias vectors, and σ is the activation function.

[0026] Optionally, the analyzing network weights includes checking the weight matrices W1, W2, and W3 of each layer in the neural network, determining the contribution of each input feature to the output score, and using a weight normalization method to normalize the absolute values ​​in the weight matrix to obtain the relative contribution of each feature: Where |W ji | represents the absolute value of the weight of the i-th feature in the j-th layer;

[0027] The feature contribution analysis includes sorting the first several features that contribute most to the health status score according to the normalized weight values, and identifying the key health indicators that affect the patient's health in combination with the importance ranking of the features.

[0028] Optionally, S3 also includes performing a baseline assessment of each key health indicator, determining the current status, and setting short-term and long-term rehabilitation goals for each key health indicator based on the patient's condition and rehabilitation needs.

[0029] Optionally, the intelligent algorithm in S4 uses a hybrid model of transfer learning and self-organizing mapping to generate a personalized rehabilitation training program, specifically including:

[0030] S41, input short-term and long-term rehabilitation goals: input the short-term and long-term rehabilitation goals for each key health indicator into the intelligent algorithm;

[0031] S42, using the pre-trained model, transferring the learning experience in similar rehabilitation tasks to the rehabilitation tasks of the current patient, and adjusting the parameters of the pre-trained model to adapt it to the specific health status and rehabilitation goals of the current patient;

[0032] S43, clustering and visualizing the data processed by transfer learning using self-organizing maps, identifying the health status patterns of patients, mapping the patient data to different health status groups through clustering, identifying similar health status patterns, and assigning the patients to the most similar health status groups through the self-organizing map clustering results to optimize the rehabilitation plan;

[0033] S44, generating a personalized rehabilitation training program.

[0034] Optionally, the S44 specifically includes:

[0035] S441, Exercise training: Develop daily and weekly exercise training plans based on targets for EMG and HRV.

[0036] S442, Physical therapy plan: Develop a personalized physical therapy plan based on self-reported pain scores and galvanic skin response targets.

[0037] S443, Psychological counseling program: Develop a psychological counseling plan based on the goals of brain waves and emotional states.

[0038] S444, Nutritional recommendations: Develop personalized nutritional recommendations based on overall health status and fatigue goals.

[0039] Optionally, the update of the transfer learning model is expressed as:

[0040] Among them, θ is the model parameter, η is the learning rate, L is the loss function, and D source is the source data, D target is the target data, and α is the adjustment coefficient.

[0041] Optionally, the update of the self-organizing map is expressed as:

[0042] w i (t+1)=w i (t)+α(t)h ci (t)[x(t)-w i (t)], where w i (t) is the weight vector of node i at time t, α(t) is the learning rate, and h ci (t) is the neighborhood function between node i and the winning node c, and x(t) is the input vector.

[0043] Beneficial effects of the present invention:

[0044] The present invention uses multimodal biofeedback data collection and fusion technology, combined with self-report data, to comprehensively evaluate the patient's health status, and uses transfer learning and self-organizing map (SOM) hybrid models to draw on effective experience from similar rehabilitation tasks, while optimizing according to the patient's specific health status pattern. This method not only improves the scientificity and accuracy of the rehabilitation program, but also can generate highly personalized rehabilitation training programs based on the specific needs and goals of each patient, including exercise training, physical therapy, psychological counseling and nutritional advice.

[0045] The present invention sets clear short-term and long-term rehabilitation goals for the identified key health indicators. The short-term goals focus on quickly improving the patient's current health status, such as improving heart rate variability, reducing pain scores, increasing muscle function, etc.; the long-term goals focus on the patient's comprehensive rehabilitation and continuous improvement of their health level. Through clear goal setting and dynamic adjustment of training plans, patients can see significant rehabilitation effects in the short term and maintain a steady improvement in their health status in the long term. This goal-oriented rehabilitation method ensures the continuity and efficiency of the rehabilitation process and provides patients with a scientific and systematic rehabilitation path.

[0046] The present invention uses a comprehensive health status assessment model, based on multimodal biofeedback data and self-report data, to systematically analyze the patient's health status and identify key health indicators (such as heart rate variability, electromyographic signals, galvanic skin response, etc.) that are crucial to the patient's rehabilitation. This method provides a clear scientific basis for the formulation of rehabilitation programs, ensuring that each step of rehabilitation measures can effectively intervene in key issues and improve the pertinence and effectiveness of rehabilitation training. In addition, by continuously monitoring these key indicators, rehabilitation progress can be dynamically evaluated, and the program can be adjusted in a timely manner to ensure the smooth realization of rehabilitation goals. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0048] Figure 1 A schematic diagram of the optimization method flow of an embodiment of the present invention;

[0049] Figure 2 It is a logical schematic diagram of an embodiment of the present invention. DETAILED DESCRIPTION

[0050] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. At the same time, it is explained here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments, and those skilled in the art may also adopt other alternatives to implement some known technologies; and the accompanying drawings are only for more specific description of the embodiments, and are not intended to specifically limit the present invention.

[0051] It should be noted that the references to "one embodiment", "an embodiment", "an exemplary embodiment", "some embodiments" and the like in the specification indicate that the embodiments described may include specific features, structures or characteristics, but not every embodiment may include the specific features, structures or characteristics. In addition, when a specific feature, structure or characteristic is described in conjunction with an embodiment, it should be within the knowledge of a person skilled in the art to implement such feature, structure or characteristic in conjunction with other embodiments (whether or not explicitly described).

[0052] In general, a term can be understood, at least in part, from its use in context. For example, depending, at least in part, on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending, at least in part, on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0053] like Figure 1-Figure 2 As shown in the figure, the biofeedback-driven personalized rehabilitation optimization method includes the following steps:

[0054] S1, data collection: collect multimodal biofeedback data from patients, including heart rate variability, galvanic skin response, electromyographic signals, breathing patterns, and brain waves, and collect self-reported data from patients through user terminal devices, including pain level, emotional state, and fatigue;

[0055] S2, multi-dimensional data fusion and preliminary analysis: pre-process the collected multi-modal biofeedback data and self-report data, use multi-dimensional data fusion algorithm to integrate data from different sources, build a comprehensive health status assessment model, evaluate the overall health status of patients based on the comprehensive health status assessment model, and identify key health indicators;

[0056] S3, personalized rehabilitation goal setting: according to the identified key health indicators, set personalized rehabilitation goals, including short-term and long-term rehabilitation goals;

[0057] S4, intelligent rehabilitation program formulation: Based on short-term and long-term rehabilitation goals, use intelligent algorithms to generate personalized rehabilitation training programs, which include sports training, physical therapy, psychological counseling, and nutritional advice.

[0058] S1 specifically includes:

[0059] S11, heart rate variability data collection: use a wearable heart rate monitoring device (heart rate belt or smart watch) to monitor the patient's heart rate data in real time, record and analyze the time interval of the heart rate, i.e., the RR interval, through the data acquisition module, and calculate the heart rate variability (HRV) index;

[0060] S12, skin electrical response data collection: Use skin potential sensors attached to the patient's palm or fingertips to measure the changes in skin conductivity in real time and record skin electrical response data to reflect the patient's emotions and stress status;

[0061] S13, electromyographic signal data acquisition: Use surface electromyography (sEMG) sensors attached to the patient's target muscle groups to record the electrical activity signals of the muscles in real time, monitor muscle contraction and relaxation, and evaluate muscle function and fatigue status;

[0062] S14, breathing pattern data acquisition: using a breathing belt to monitor the patient's breathing rate and depth, and recording breathing pattern data, including breathing rate, through a data acquisition module;

[0063] S15, EEG data collection: Use a portable electroencephalogram (EEG) device to monitor the patient's brain wave activity, record brain wave signals in real time, and analyze EEG activities in different frequency bands;

[0064] S16, self-report data collection: A user interface is provided through the user terminal device (smartphone or tablet computer), through which the patient enters the self-report data, and regular reminders are set to remind the patient to fill in the pain score, emotional state and fatigue data.

[0065] S2 specifically includes:

[0066] S21, data preprocessing: filtering technology was used to remove noise from heart rate variability, galvanic skin response, electromyographic signals, breathing patterns, and brain wave data. Each biofeedback data was standardized so that different types of data had the same dimension and range. The z-score standardization method was used to convert the data into a standard normal distribution with a mean of 0 and a standard deviation of 1.

[0067] S22, using feature selection and feature extraction techniques, extract features from data from different sources, construct feature vectors, and use independent component analysis dimensionality reduction algorithm to reduce data dimensions and reduce redundant information. Let the original data matrix be X∈R m×n , where m is the number of samples and n is the number of features. Independent component analysis is used to reduce the dimension of the standardized data and extract independent components. The standardized data is expressed as: Z = [z1, z2, ..., z n ], using independent component analysis Z = AS, where A is the mixing matrix and S is the independent component matrix, and independent components are extracted by solving S;

[0068] S23, based on the multimodal fusion neural network, the standardized biofeedback data and self-report data are integrated to build a comprehensive health status assessment model to reflect the patient's overall health status. The multimodal fusion neural network takes different types of biofeedback data and self-report data as input, and the input vector is I = [i1, i2, ..., i k], where k is the number of data types. The network structure of the multimodal fusion neural network includes an input layer, multiple hidden layers, and an output layer. The input layer receives the fused feature vector, the hidden layer performs nonlinear transformation, and the output layer generates a health status score. ;

[0069] S24, health status assessment: input the real-time collected biofeedback data and self-report data into the comprehensive health status assessment model, and based on the model output results, evaluate the patient's overall health status and generate a health status score;

[0070] S25, Identify key health indicators: By analyzing network weights and feature contribution analysis, identify key health indicators that affect patients' health. Key health indicators include heart rate variability, electromyographic signal strength, galvanic skin response, electromyographic signal, respiratory rate, brain waves, self-reported pain level, self-reported emotional state, and self-reported fatigue.

[0071] Comprehensive health status score (i.e., the comprehensive health status assessment model expression) is expressed as: Where W1, W2, W3 are weight matrices, b1, b2, b3 are bias vectors, and σ is the activation function.

[0072] Analyzing network weights includes looking at the weight matrices W1, W2, and W3 of each layer in the neural network, determining the contribution of each input feature to the output score, and using the weight normalization method to normalize the absolute values ​​in the weight matrix to obtain the relative contribution of each feature: Where |W ji | represents the absolute value of the weight of the i-th feature in the j-th layer;

[0073] Feature contribution analysis includes sorting the top features that contribute most to the health status score according to the normalized weight values, and identifying key health indicators that affect the patient's health based on the feature importance ranking.

[0074] S3 also includes conducting a baseline assessment of each key health indicator, determining the current status, and setting short-term and long-term rehabilitation goals for each key health indicator based on the patient's condition and rehabilitation needs;

[0075] Conduct a baseline assessment of each key health indicator to determine the current status:

[0076] i. Heart rate variability (HRV):

[0077] The heart rate data of the patients were continuously recorded using a heart rate monitoring device, the RR interval was calculated, and the standard deviation (SDNN), root mean square (RMSSD) and low frequency / high frequency ratio (LF / HF) of HRV were analyzed.

[0078] ii.Galvanic Skin Response (GSR):

[0079] Use a galvanic skin response sensor to measure the patient's skin conductivity in different states, record and analyze changes in galvanic skin response, and assess baseline galvanic skin response levels.

[0080] iii. Electromyography (EMG):

[0081] A surface electromyography (sEMG) sensor was used to record the electrical activity signals of specific muscle groups of the patient, and the root mean square value (RMS) and mean frequency (MPF) of the EMG signals were analyzed to determine the baseline state of muscle function.

[0082] iv.Breathing Pattern:

[0083] Record the patient's respiratory rate using a respiratory monitoring device; analyze the stability and regularity of the respiratory rate to determine baseline respiratory status.

[0084] v.Electroencephalogram (EEG):

[0085] A portable electroencephalogram (EEG) device was used to record the patient's brain wave activity, analyze the intensity and distribution of alpha brain waves in different frequency bands, and determine the baseline brain wave state.

[0086] vi. Self-reported pain scores:

[0087] Patients were asked to report their pain intensity using a standardized pain rating scale (such as VAS), and the baseline value of the pain score was recorded.

[0088] vii. Self-reported emotional state:

[0089] The mood state of the patients was assessed using the mood state scale (Beck Depression Inventory), and the baseline score of the mood state was recorded.

[0090] viii. Self-reported fatigue: patients were asked to report their fatigue level using a fatigue severity scale (Fatigue Severity Scale (FSS)) and the baseline value of fatigue was recorded.

[0091] Examples of short-term rehabilitation goal setting include:

[0092] i. Heart rate variability (HRV): Improve HRV by 5% in one month by doing 20 minutes of aerobic exercise and deep breathing exercises every day;

[0093] ii. Galvanic Skin Response (GSR): Reducing GSR levels within two weeks through 30 minutes of daily meditation and relaxation training;

[0094] iii. Electromyographic signal (EMG): Improve the electromyographic signal strength of specific muscle groups by 10% within one month by performing specific strength training;

[0095] iv. Breathing pattern: Achieve a stable breathing rate and depth within two weeks through 15 minutes of daily breathing training;

[0096] v. EEG: Increase alpha wave activity within one month through EEG feedback training 3 times a week;

[0097] vi. Self-reported pain score: Reduce pain score by 2 points within 2 weeks through regular physical therapy and moderate exercise;

[0098] vii. Self-reported emotional state: Improved emotional state within one month through weekly counseling and relaxation training;

[0099] viii. Self-reported fatigue: Reduce fatigue within two weeks by adjusting your sleep schedule and increasing rest time.

[0100] c. Examples of long-term rehabilitation goal setting are as follows:

[0101] i. Heart rate variability (HRV): restore HRV to the average level of a healthy population within six months through continued aerobic exercise and stress management;

[0102] ii. Galvanic Skin Response (GSR): Keeping the galvanic skin response level within the normal range for three months through long-term meditation and mental relaxation exercises;

[0103] iii. Electromyographic signal (EMG): Improve whole body muscle function, especially the strength and endurance of specific muscle groups, through systematic strength and endurance training within six months;

[0104] iv. Breathing pattern: within three months, the breathing rate and depth should reach a healthy level through long-term breathing training and aerobic exercise;

[0105] v. EEG: Significantly increased alpha and theta wave activity over six months through regular EEG feedback and cognitive training;

[0106] vi. Self-reported pain scores: Elimination or significant reduction of pain within three months with continued physical therapy, medication management, and moderate exercise;

[0107] vii. Self-reported emotional state: Maintaining a positive emotional state over a six-month period through long-term psychological counseling and lifestyle adjustments;

[0108] viii. Self-reported fatigue: Maintain low fatigue levels over three months by optimizing sleep and rest schedules, increasing rest time, and exercising as needed.

[0109] The intelligent algorithm in S4 uses a hybrid model of transfer learning and self-organizing maps to generate personalized rehabilitation training programs, including:

[0110] S41, Input short-term and long-term recovery goals: Input short-term and long-term recovery goals for each key health indicator into the intelligent algorithm, including goals for heart rate variability (HRV), galvanic skin response (GSR), electromyographic signal (EMG), breathing pattern, brain wave (EEG), self-reported pain score, emotional state and fatigue;

[0111] S42, using the pre-trained model, transfer the learning experience in similar rehabilitation tasks to the rehabilitation task of the current patient, adjust the parameters of the pre-trained model to adapt it to the specific health status and rehabilitation goals of the current patient, and the update of the transfer learning model is expressed as:

[0112] Among them, θ is the model parameter, η is the learning rate, L is the loss function, and D source is the source data, D target is the target data, and α is the adjustment coefficient. This method enables the model to quickly adapt to new tasks using existing experience;

[0113] S43, clustering and visualizing the data processed by transfer learning using self-organizing maps, identifying the health status patterns of patients, mapping patient data to different health status groups through clustering, identifying similar health status patterns, and assigning patients to the most similar health status groups through self-organizing map clustering results to optimize rehabilitation plans. The update of the self-organizing map is expressed as:

[0114] w i (t+1)=w i (t)+α(t)h ci (t)[x(t)-w i (t)], where w i (t) is the weight vector of node i at time t, α(t) is the learning rate, and h ci (t) is the neighborhood function between node i and the winning node c, and x(t) is the input vector;

[0115] S44, generating a personalized rehabilitation training program.

[0116] S44 specifically includes:

[0117] S441, Exercise training: Based on the goals of electromyographic signals and heart rate variability, develop daily and weekly exercise training plans, including aerobic exercise, strength training, and flexibility training. Use transfer learning to preliminarily adjust the intensity and type of training based on the patient's real-time feedback, and further optimize the training plan based on the clustering results of the self-organizing map. For example, if the clustering results of the self-organizing map show that the heart rate variability of a certain type of patient is significantly improved after strength training, then similar strength training content will be added to the plan for the current patient.

[0118] S442, Physical therapy plan: Based on the self-reported pain score and skin electrical response target, develop a personalized physical therapy plan, such as heat compress, massage, and electrical stimulation. Use transfer learning to generate a preliminary physical therapy plan, and optimize the physical therapy time and frequency in combination with the self-organizing map results to ensure the effectiveness of the physical therapy plan. For example, if the self-organizing map clustering results show that a certain type of patient responds well to electrical stimulation, increase the frequency and time of electrical stimulation physical therapy.

[0119] S443, psychological counseling plan: Based on the goals of brain waves and emotional states, a psychological counseling plan is formulated, including psychological counseling, cognitive behavioral therapy, meditation and relaxation training. A preliminary psychological counseling plan is generated by combining the transfer learning model, and the self-organizing map results are dynamically adjusted to improve the effect of psychological counseling. For example, if the self-organizing map clustering shows that meditation has a significant effect on improving the emotional state of a certain type of patient, the frequency and time of meditation training will be increased.

[0120] S444, Nutritional advice: Based on the overall health status and fatigue goals, formulate personalized nutritional advice, including dietary structure and nutritional supplements, use transfer learning to provide preliminary nutritional advice, and combine self-organizing map analysis for further optimization to ensure the scientific nature and personalization of nutritional advice. For example, if the self-organizing map clustering results show that a certain type of patient responds well to a high-protein diet, then the recommendation for a high-protein diet will be increased.

[0121] The present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention. In order to make the public have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, but those skilled in the art can fully understand the present invention without the description of these details. In addition, in order to avoid unnecessary confusion about the essence of the present invention, well-known methods, processes, procedures, components and circuits are not described in detail.

[0122] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A biofeedback-driven personalized rehabilitation optimization method, characterized in that: The following steps are involved: S1, data collection: collect multimodal biofeedback data from patients, including heart rate variability, galvanic skin response, electromyographic signals, breathing patterns, and brain waves, and collect self-reported data from patients through user terminal devices, including pain level, emotional state, and fatigue; S2, multi-dimensional data fusion and preliminary analysis: pre-process the collected multi-modal biofeedback data and self-report data, use multi-dimensional data fusion algorithm to integrate data from different sources, build a comprehensive health status assessment model, evaluate the overall health status of patients based on the comprehensive health status assessment model, and identify key health indicators; S3, personalized rehabilitation goal setting: according to the identified key health indicators, set personalized rehabilitation goals, including short-term and long-term rehabilitation goals; S4, intelligent rehabilitation program formulation: Based on short-term and long-term rehabilitation goals, use intelligent algorithms to generate personalized rehabilitation training programs, which include sports training, physical therapy, psychological counseling, and nutritional advice.

2. The biofeedback-driven personalized rehabilitation optimization method according to claim 1, characterized in that: The S1 specifically includes: S11, heart rate variability data collection: using a wearable heart rate monitoring device to monitor the patient's heart rate data in real time, recording and analyzing the heart rate time interval through a data collection module, and calculating the heart rate variability index; S12, skin electrical response data collection: Use a skin potential sensor attached to the patient's palm or fingertips to measure the conductivity changes of the skin in real time and record the skin electrical response data; S13, EMG data acquisition: Use surface EMG sensors attached to the patient's target muscle groups to record the muscle's electrical activity signals in real time, monitor muscle contraction and relaxation, and evaluate muscle function and fatigue status; S14, breathing pattern data acquisition: using a breathing belt to monitor the patient's breathing rate and depth, and recording breathing pattern data, including breathing rate, through a data acquisition module; S15, EEG data collection: Use portable EEG equipment to monitor the patient's brain wave activity, record brain wave signals in real time, and analyze EEG activities in different frequency bands; S16, self-report data collection: a user interface is provided through the user terminal device, through which the patient inputs the self-report data, and regular reminders are set to remind the patient to fill in the pain score, emotional state and fatigue data.

3. The biofeedback-driven personalized rehabilitation optimization method according to claim 1, characterized in that: The S2 specifically includes: S21, data preprocessing: filtering technology was used to remove noise from heart rate variability, galvanic skin response, electromyographic signals, breathing patterns, and brain wave data. Each biofeedback data was standardized so that different types of data had the same dimension and range. The z-score standardization method was used to convert the data into a standard normal distribution with a mean of 0 and a standard deviation of 1. S22, using feature selection and feature extraction techniques, extract features from data from different sources, construct feature vectors, and use independent component analysis dimensionality reduction algorithm to reduce data dimensions and reduce redundant information. Let the original data matrix be X∈R m×n , where m is the number of samples and n is the number of features. Independent component analysis is used to reduce the dimension of the standardized data and extract independent components. The standardized data is expressed as: Z = [z1, z2, ..., z n ], using independent component analysis Z = AS, where A is the mixing matrix and S is the independent component matrix, and independent components are extracted by solving S; S23, based on a multimodal fusion neural network, the standardized biofeedback data and self-report data are integrated to construct a comprehensive health status assessment model to reflect the overall health status of the patient. The multimodal fusion neural network takes different types of biofeedback data and self-report data as input, and the input vector is I = [i1, i2, ..., i k ], where k is the number of data types. The network structure of the multimodal fusion neural network includes an input layer, multiple hidden layers, and an output layer. The input layer receives the fused feature vector, the hidden layer performs nonlinear transformation, and the output layer generates a health status score. S24, health status assessment: input the real-time collected biofeedback data and self-report data into the comprehensive health status assessment model, and based on the model output results, evaluate the patient's overall health status and generate a health status score; S25, Identify key health indicators: By analyzing network weights and feature contribution analysis, identify key health indicators that affect patients' health. Key health indicators include heart rate variability, electromyographic signal strength, galvanic skin response, electromyographic signal, respiratory rate, brain waves, self-reported pain level, self-reported emotional state, and self-reported fatigue.

4. The biofeedback-driven personalized rehabilitation optimization method according to claim 3, characterized in that: The comprehensive health status score It is expressed as: Where W1, W2, W3 are weight matrices, b1, b2, b3 are bias vectors, and σ is the activation function.

5. The biofeedback-driven personalized rehabilitation optimization method according to claim 3, characterized in that: The analysis of network weights includes checking the weight matrices W1, W2, and W3 of each layer in the neural network, determining the contribution of each input feature to the output score, and using the weight normalization method to normalize the absolute values ​​in the weight matrix to obtain the relative contribution of each feature: Contribution Where |W ji | represents the absolute value of the weight of the i-th feature in the j-th layer; The feature contribution analysis includes sorting the first several features that contribute most to the health status score according to the normalized weight values, and identifying the key health indicators that affect the patient's health in combination with the importance ranking of the features.

6. The biofeedback-driven personalized rehabilitation optimization method according to claim 5, characterized in that: The S3 also includes conducting a baseline assessment of each key health indicator, determining the current status, and setting short-term and long-term rehabilitation goals for each key health indicator based on the patient's condition and rehabilitation needs.

7. The biofeedback-driven personalized rehabilitation optimization method according to claim 1, characterized in that: The intelligent algorithm in S4 uses a hybrid model of transfer learning and self-organizing mapping to generate a personalized rehabilitation training program, specifically including: S41, input short-term and long-term rehabilitation goals: input the short-term and long-term rehabilitation goals for each key health indicator into the intelligent algorithm; S42, using the pre-trained model, transferring the learning experience in the rehabilitation task to the rehabilitation task of the current patient, and adjusting the parameters of the pre-trained model to adapt it to the specific health condition and rehabilitation goal of the current patient; S43, clustering and visualizing the data processed by transfer learning using self-organizing maps, identifying the health status patterns of patients, mapping the patient data to different health status groups through clustering, identifying similar health status patterns, and assigning the patients to the most similar health status groups through the self-organizing map clustering results to optimize the rehabilitation plan; S44, generating a personalized rehabilitation training program.

8. The biofeedback-driven personalized rehabilitation optimization method according to claim 7, characterized in that: The S44 specifically includes: S441, Exercise training: Develop daily and weekly exercise training plans based on targets for EMG and HRV; S442, Physical therapy plan: Develop a personalized physical therapy plan based on self-reported pain scores and galvanic skin response targets; S443, Psychological counseling program: Develop a psychological counseling plan based on the goals of brain waves and emotional states; S444, Nutritional recommendations: Develop personalized nutritional recommendations based on overall health status and fatigue goals.

9. The biofeedback-driven personalized rehabilitation optimization method according to claim 7, characterized in that: The update of the transfer learning model is expressed as: Among them, θ is the model parameter, L is the learning rate, L is the loss function, and D source is the source data, D target is the target data, and α is the adjustment coefficient.

10. The biofeedback-driven personalized rehabilitation optimization method according to claim 7, characterized in that: The update of the self-organizing map is expressed as: w i (t+1)=w i (y)+α(t)h ci (t)[x(t)-w i (t)], where w i (t) is the weight vector of node i at time t, α(t) is the learning rate, and h ci (t) is the neighborhood function between node i and the winning node c, and x(t) is the input vector.

Citation Information

Patent Citations

  • Senile dementia monitoring system based on healthy service robot

    CN105078449A

  • Emotion recognition method based on adaptive fuzzy support vector machine

    CN108309328A

  • Multi-dimensional intervention chronic disease lifestyle medical management system

    CN110491526A

  • Myoelectricity and muscle force prediction method suitable for hand function rehabilitation training of children with cerebral palsy

    CN116869535A

  • Dynamic adaptive diet suggestion generation method based on real-time biomarker monitoring

    CN118116550A

Cited By

  • Stroke rehabilitation training system and method based on deep reinforcement learning

    CN120072195A

  • Exercise rehabilitation effect and range quantitative evaluation method and system

    CN120089381A

  • Rheumatism rehabilitation intelligent evaluation system based on ICF idea and classification system

    CN120708918A

  • Rehabilitation intelligent assessment system for rheumatism based on ICF concept and classification system

    CN120708918B

  • Rehabilitation training interaction task setting method and system based on medical treatment and experience double guidance

    CN121148592A