Ventilation-oxygenation dynamic balance guided respiratory rehabilitation training system and control method

Abstract

The application discloses a ventilation-oxygenation dynamic balance guided respiratory rehabilitation training system and a control method, belongs to the technical field of respiratory rehabilitation training equipment, and comprises a user and signal input layer, an intelligent control and decision layer and an execution and output layer, wherein the user and signal input layer comprises a multi-parameter physiological monitoring module, a man-machine interaction module and a scheme configuration module, the matched hardware components of the user and signal input layer include a breathing mouthpiece, a breathing pipeline, a pulse blood oxygen saturation sensor and an intelligent terminal, the intelligent control and decision layer comprises a signal processing and state analysis module, an individualized impedance and respiratory mode training module and a master control module, the execution and output layer comprises an adjustable impedance execution module and an oxygen supply module, and the matched hardware components of the intelligent control and decision layer and the execution and output layer are a host computer. The application integrates the pulse blood oxygen saturation sensor and the end-tidal carbon dioxide partial pressure sensor, acquires physiological signals related to training, and combines the oxygen concentration adjustable supply module to provide basic guarantee for training safety.

Description

Technical Field

[0001] This invention belongs to the technical field of respiratory rehabilitation training equipment, specifically relating to a ventilation-oxygenation dynamic balance-guided respiratory rehabilitation training system and control method. Background Technology

[0002] Respiratory muscle training effectively improves breathing difficulties, enhances exercise endurance, and improves quality of life, forming the cornerstone of pulmonary rehabilitation. Traditional respiratory muscle training methods, such as pursed-lip breathing and diaphragmatic breathing, can effectively improve lung function and exercise capacity. However, these methods have limitations in providing users with reasonable, individualized, and precise rehabilitation plans. In recent years, resistance-based respiratory muscle training has gradually become the main method of clinical respiratory muscle training. There are three main types of resistance-based respiratory muscle training: aperture resistance loading, threshold loading, and cone resistance loading. Aperture resistance loading controls inspiratory resistance by changing the inspiratory orifice diameter; its load magnitude changes exponentially with the orifice diameter, making it difficult to accurately adjust the resistance. Threshold loading is currently the mainstream clinical application; its resistance magnitude is constant and does not change with airflow, providing users with precise respiratory muscle strength training. However, its "fixed load" mode does not match the "strength-length relationship" of respiratory muscle physiology, leading to limited tidal volume and affecting training effectiveness and compliance. The cone-shaped flow resistance load curve dynamically reduces resistance during inspiration based on changes in lung volume and respiratory muscle strength, thus better reflecting the pressure-volume relationship of the respiratory system. Compared to threshold loading, this loading mode better matches the physiological characteristics of respiratory muscles, generating greater tidal volume and respiratory work, and has unique advantages in improving ventilation and respiratory endurance. Resistance training of respiratory muscles is essentially a type of load training, which strengthens the inspiratory and expiratory muscles by increasing resistance during inspiration or expiration. However, in clinical use, resistance trainers can only train muscle strength and endurance, neglecting the user's physiological state. Therefore, it is easier to induce respiratory muscle fatigue, exercise-induced hypoxia, and hypercapnia during training, causing acid-base imbalance in the respiratory system and leading to safety issues.

[0003] Therefore, there is an urgent need for a respiratory rehabilitation training system and control method that can integrate multi-parameter physiological monitoring, realize individualized dynamic impedance regulation, and provide physiological parameter feedback, and guide the dynamic balance of ventilation and oxygenation. Summary of the Invention

[0004] The purpose of this invention is to provide a ventilation-oxygenation dynamic balance-guided respiratory rehabilitation training system and control method, which can integrate multi-parameter physiological monitoring, realize individualized dynamic impedance regulation, and provide physiological parameter feedback, so as to solve the three major problems of training safety, effectiveness and compliance, and form a new mode of respiratory muscle training.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] A ventilation-oxygenation dynamic balance-guided respiratory rehabilitation training system includes a user and signal input layer, an intelligent control and decision-making layer, and an execution and output layer. The user and signal input layer includes a multi-parameter physiological monitoring module, a human-computer interaction module, and a protocol configuration module. The supporting hardware components of the user and signal input layer include a breathing mouthpiece, breathing tubing, a pulse oximetry sensor, and a smart terminal. The intelligent control and decision-making layer includes a signal processing and status analysis module, an individualized impedance and breathing pattern training module, and a main control module. The execution and output layer includes an adjustable impedance execution module and an oxygen supply module. The supporting hardware components of the intelligent control and decision-making layer and the execution and output layer are the host unit.

[0007] A further improvement of the technical solution of the present invention is that: the host includes a main control circuit board with an integrated microprocessor, an adjustable impedance valve, an oxygen supply structure, a power supply module and a status display screen; the signal processing and status analysis module of the intelligent control and decision layer, the individualized impedance and breathing pattern training module and the main control module are all integrated on the microprocessor of the main control circuit board; and the adjustable impedance execution module and the oxygen supply module of the execution and output layer are respectively integrated on the adjustable impedance valve and the oxygen supply structure.

[0008] A further improvement of the technical solution of this invention is as follows: the adjustable impedance valve is driven by a stepper motor to rotate the valve, providing precise training resistance by changing the orifice diameter of the airflow channel; the oxygen supply structure is controlled by an oxygen supply flow regulating valve to control the oxygen supply; the breathing mouthpiece is connected to the internal airway interface of the host through a breathing tube, which integrates a pressure sensor and an end-expiratory carbon dioxide partial pressure sensor for collecting the user's respiratory pressure, respiratory flow, tidal volume, and end-expiratory carbon dioxide partial pressure; the pulse oximetry sensor adopts any one of the finger clip, patch, or wrist type, and is connected to the signal acquisition interface of the host through wired or wireless means for real-time monitoring of the user's blood oxygen-related physiological parameters; the smart terminal is a tablet computer or mobile phone, which establishes a data connection with the main control circuit board of the host through Wi-Fi or Bluetooth to receive data, display real-time curves, provide training guidance and alarms, and complete the training program configuration.

[0009] A method for controlling respiratory rehabilitation training guided by dynamic ventilation-oxygenation balance, using a respiratory rehabilitation training system, includes the following steps:

[0010] Step S1: Initialization and Signal Acquisition: Training begins. The user-signal input layer's human-computer interaction module loads the initialization training prescription. The user-signal input layer's multi-parameter physiological monitoring module, relying on the breathing tubing and pulse oximetry sensor, acquires pressure signal P and blood oxygen saturation. and end-tidal carbon dioxide partial pressure ;

[0011] Step S2: Signal Processing: The signal processing and state analysis module of the intelligent control and decision-making layer calculates the flow rate Q and tidal volume in real time based on the pressure signal P. Inhalation time and exhalation time ;

[0012] Step S3: Safety threshold judgment: Pre-set the safety judgment thresholds for each physiological parameter, compare the monitoring data with the threshold standards in real time, and immediately start the safety intervention sub-process when the parameter deviates from the safe range: First, determine whether the blood oxygen related parameters are abnormal, and then classify and deal with them in combination with the end-tidal carbon dioxide partial pressure value. According to different abnormal combinations, take corresponding measures such as stopping resistance training, adjusting oxygen supply concentration, and guiding standardized breathing mode to ensure that the training process is safe and controllable.

[0013] Step S4: When all physiological parameters are within the safe range, the individualized impedance and breathing pattern training module of the intelligent control and decision-making layer executes the individualized impedance and breathing pattern optimization algorithm.

[0014] Step S5: Execution and Output: The main control module of the intelligent control and decision-making layer transmits the optimized instructions to the adjustable impedance execution module and oxygen supply module of the execution and output layer, which are then converted into execution operations by the stepper motor, oxygen supply flow regulating valve, and voice prompts.

[0015] Step S6: Loop and End: Repeat steps S1-S5 to perform signal acquisition, processing, judgment and individualized training parameter optimization for the next cycle until the training duration is reached, then end the training and the smart terminal 5 generates a training report.

[0016] A further improvement to the technical solution of this invention lies in the following: In step S4, the individualized impedance and breathing pattern training module of the intelligent control and decision-making layer executes the individualized impedance and breathing pattern optimization algorithm, including the following steps:

[0017] Step S401: Model building;

[0018] Step S402: Parameter identification and calculation;

[0019] Step S403: Construct the MPC-BO optimization problem;

[0020] Step S404: Call the model;

[0021] Step S405: Evaluate the cost.

[0022] A further improvement of the technical solution of the present invention is that: in step S401, a pre-model of respiratory mode and ventilation oxygenation demand is established based on the respiratory system motion equation, alveolar gas equation, and blood oxygenation dissociation curve equation.

[0023] The equation of motion for the respiratory system is expressed by the following formula:

[0024]

[0025]

[0026] In the formula, Indicates oral pressure; Indicates the force exerted by the respiratory muscles; Indicates the elasticity of the respiratory system; Indicates airway resistance; Indicates tidal volume; This indicates the resistance generated by the respiratory muscle trainer; Indicates positive end-expiratory pressure; Indicates alveolar ventilation; Indicates tidal volume; Indicates the volume of dead space;

[0027] It can be expressed by the following formula:

[0028]

[0029] In the formula, Indicates the amplitude of inspiratory muscle exertion; The time constant representing the exponential expiratory profile. ; Indicates the inhalation time; Indicates the exhalation time; Indicates the respiratory cycle;

[0030] The alveolar gas equation is expressed by the following formula:

[0031]

[0032] In the formula, Indicates alveolar oxygen partial pressure; Indicates the partial pressure of carbon dioxide in the alveoli; Indicates the fraction of inhaled oxygen concentration; Indicates atmospheric pressure; Indicates the partial pressure of water vapor; Indicates the respiratory exchange ratio;

[0033] The equation for the blood oxygenation dissociation curve is expressed by the following formula:

[0034]

[0035]

[0036]

[0037]

[0038]

[0039]

[0040]

[0041]

[0042] In the formula, Indicates blood oxygen saturation; Represents the Hill coefficient; This represents the partial pressure of oxygen when blood oxygen saturation is 50%. , , This represents the model's intrinsic parameters, obtained by fitting data from normal blood under standard physiological conditions. =2.82, =1.20, =29.25 mmHg; Indicates blood concentration; This indicates the concentration of hydrogen ions in the blood; This represents the standard half-saturated oxygen partial pressure, with a value of 26.8 mmHg; This represents the standard pH value, which is 7.4. This represents the standard CO2 partial pressure, with a value of 40 mmHg. It represents 2,3-diphosphoglyceric acid; This indicates the standard 2,3-DPG concentration, with values ​​ranging from [value missing]. M (mol / L); Indicates blood temperature; This represents the standard blood temperature, and its value is... .

[0043] A further improvement to the technical solution of this invention lies in: in step S402, the user is allowed to breathe in a calm state for a period of time, and the user's respiratory system resistance R and elasticity E are identified using data on oral pressure, flow rate, and tidal volume. The Levenberg-Marquardt algorithm is then used to... Fit the model to make it conform to individual characteristics;

[0044] In the given discrete time points Data set obtained from the above measurement Estimate parameter vector ;

[0045] The objective function can be expressed as:

[0046]

[0047] In the formula, Indicates the first The residual of each data point is the difference between the measured stress and the model-predicted stress.

[0048] The residual can be defined as:

[0049] ;

[0050] Define residual vector Then the Jacobian matrix can be expressed as:

[0051]

[0052] The LM iterative equation can be expressed as: ,in .

[0053] A further improvement of the technical solution of the present invention is as follows: In step S403, an individualized impedance and breathing pattern optimization algorithm MPC-BO is proposed based on the Model Predictive Control (MPC) framework and using Bayesian Optimization (BO) as the solver. After each breathing training session, the algorithm continuously optimizes the equipment control parameters for the next training session based on the respiratory-gas exchange prediction model constructed according to the physiological mechanism, transforming the clinical rehabilitation goal into a finite-time domain optimization control problem, and using BO to find the solution to this nonlinear, non-convex optimization problem.

[0054] The general form of the nonlinear MPC state-space equation of the system is:

[0055]

[0056] Wherein, the state vector Control input vector Output vector With interference vector They are represented as follows:

[0057] , , , ;

[0058] In the formula, Indicates the rate of carbon dioxide production;

[0059] Equations of state Represented as :

[0060]

[0061] In the formula, Represents the standard conversion constant;

[0062] Output equation Represented as :

[0063]

[0064] The forward Euler method is used to sample the continuous model at a fixed sampling period. Discretization yields the discrete-time prediction model:

[0065]

[0066] At the decision point after each training session, i.e. after the Tth training session, MPC is based on the current state. Solve a finite-time open-loop optimization problem to obtain the optimal control command for the next training iteration, i.e., the optimal control command for the (T+1)th training iteration.

[0067] The optimization objective is defined as minimizing the deviation between the physiological indicators and the target values ​​within the prediction time domain, while also considering control smoothness. Considering the periodic independence of respiratory muscle training (i.e., each cycle resets the state variables at the end of exhalation), and setting both the prediction and control time domains to 1, the cost function is designed as follows:

[0068]

[0069] In the formula, , These represent the upper and lower limits of the clinically safe range for blood oxygen saturation, respectively. , These represent the upper and lower limits of the clinically safe range for the partial pressure of carbon dioxide at the end of expiration, respectively. , These represent the predicted values ​​of blood oxygen saturation and carbon dioxide partial pressure at the end of the respiratory tract, respectively. , , This represents the weighting coefficient, used to measure the importance of different objectives; This indicates the control value from the previous cycle, preventing drastic changes in the control value and avoiding user discomfort.

[0070] BO solver in control variables Several sets are randomly generated within the feasible region, and the corresponding costs are obtained through the cost function. and form a sample point dataset. The distribution of the objective function is learned from the evaluated sample point dataset using a Gaussian process as a probabilistic surrogate model; the unknown cost function is assumed. Follows a Gaussian process:

[0071]

[0072] In the formula, This represents the mean function, and is usually set to a constant or zero; This represents the covariance function, also known as the kernel function, used to measure the input points. and Similarity between them;

[0073] The Matrn 5 / 2 kernel function is selected because it can smooth the physiological response surface curve.

[0074]

[0075] In the formula, This represents the Euclidean distance between two points in the input space. It represents the signal variance, indicating the overall fluctuation range of the control output value; The length dimension represents the rate at which the correlation between input features decays.

[0076] In obtaining the observation dataset Then, for a new candidate point Its function value The posterior distribution is still a Gaussian distribution:

[0077] posterior mean It is the best estimate of the function value, posterior variance. Quantify the uncertainty of this estimate; the posterior mean and variance can be calculated using the following formulas:

[0078]

[0079]

[0080] In the formula, This represents the covariance matrix of the observed points; This represents the covariance vector between the new observation point and the previously observed points; Represents the observed cost vector; Indicates the variance of observation noise;

[0081] BO Maximization Acquisition Function To determine the next assessment point This function balances the balance between Small area search, that is, developing the currently known optimal point, in Large-scale regional search relationships, i.e., exploring unknown areas to reduce uncertainty;

[0082] Using the expected improvement as the acquisition function, the current optimal observation value is defined as... The desired improved function is:

[0083]

[0084] In the formula, ; and Let represent the cumulative distribution function and probability density function of the standard normal distribution, respectively.

[0085] A further improvement to the technical solution of the present invention lies in: in step S404, using This is the initial state. Simulate control commands to calculate the actual cost. And update the dataset .

[0086] A further improvement of the technical solution of the present invention lies in: in step S405, selecting the control parameter combination with the lowest cost and outputting the optimal control command. .

[0087] The technological advancements achieved by this invention due to the adoption of the above technical solutions are as follows:

[0088] The ventilation-oxygenation dynamic balance-guided respiratory rehabilitation training system and control method provided by this invention acquires training-related physiological signals by integrating a pulse oximetry sensor and an end-expiratory carbon dioxide partial pressure sensor, and provides a basic guarantee for training safety by combining an adjustable oxygen supply module.

[0089] This invention uses an individualized impedance and breathing pattern optimization algorithm to dynamically adjust the impedance parameters output by the device and provide adaptive breathing rhythm guidance, thereby improving the effectiveness of training and user compliance.

[0090] This invention integrates a pulse oximetry sensor and an end-tidal carbon dioxide partial pressure sensor. Based on the collected physiological parameters, it can automatically issue warnings when it detects a decrease in blood oxygen saturation or a trend of acid-base imbalance in the respiratory system, ensuring training safety. At the same time, through an individualized impedance and breathing pattern training module, it dynamically adjusts the device's output impedance, breathing guidance mode, and oxygen concentration of the mixed gas based on feedback from physiological parameters during training, improving training safety, effectiveness, and compliance, thus forming a new paradigm for respiratory muscle training. Attached Figure Description

[0091] Figure 1 This is a schematic diagram of the system structure of the present invention;

[0092] Figure 2 This is a schematic diagram of the system functional modules of the present invention;

[0093] Figure 3 This is a system control flowchart of the present invention;

[0094] Figure 4 This is a flowchart of the individualized impedance and breathing pattern optimization algorithm of the present invention;

[0095] The components include: 1. Breathing mouthpiece, 2. Breathing tubing, 3. Main unit, 4. Pulse oxygen saturation sensor, and 5. Smart terminal. Detailed Implementation

[0096] The present invention will be further described in detail below with reference to embodiments:

[0097] Example 1

[0098] like Figure 1 and Figure 2 As shown, this embodiment provides a ventilation-oxygenation dynamic balance-guided respiratory rehabilitation training system, including a user and signal input layer, an intelligent control and decision-making layer, and an execution and output layer. The user and signal input layer includes a multi-parameter physiological monitoring module, a human-computer interaction module, and a program configuration module. The supporting hardware components of the user and signal input layer include a breathing mouthpiece 1, a breathing tubing 2, a pulse oximetry sensor 4, and a smart terminal 5. The intelligent control and decision-making layer includes a signal processing and status analysis module, an individualized impedance and breathing pattern training module, and a main control module. The execution and output layer includes an adjustable impedance execution module and an oxygen supply module. The supporting hardware component of the intelligent control and decision-making layer and the execution and output layer is the host 3.

[0099] The host unit 3 includes a main control circuit board with an integrated microprocessor, an adjustable impedance valve, an oxygen supply structure, a power module, and a status display screen. The signal processing and status analysis module for the intelligent control and decision-making layer, the individualized impedance and breathing pattern training module, and the main control module are all integrated on the microprocessor of the main control circuit board. The adjustable impedance execution module and the oxygen supply module for the execution and output layer are integrated on the adjustable impedance valve and the oxygen supply structure, respectively. Specifically, the adjustable impedance valve is driven by a stepper motor to rotate the valve, providing precise training resistance by changing the orifice diameter of the airflow channel. The oxygen supply structure controls the oxygen supply amount through an oxygen flow regulating valve.

[0100] Furthermore, the mouthpiece 1 is connected to the internal airway interface of the main unit 3 via the breathing tubing 2. The breathing tubing 2 integrates a pressure sensor and an end-expiratory carbon dioxide partial pressure sensor to collect the user's respiratory pressure, respiratory flow, tidal volume, and end-expiratory carbon dioxide partial pressure. The pulse oximetry sensor 4 adopts any one of the following: finger clip, patch, or wrist type. It is connected to the signal acquisition interface of the main unit 3 via wired or wireless connection to monitor the user's blood oxygen-related physiological parameters in real time. The smart terminal 5 is a tablet or mobile phone with a dedicated APP installed. It establishes a data connection with the main control circuit board of the main unit 3 via Wi-Fi or Bluetooth to receive data, display real-time curves, provide training guidance and alarms, and complete the training program configuration.

[0101] During respiratory muscle training, the multi-parameter physiological monitoring module acquires the user's physiological signals using the pulse oximetry sensor 4 and the end-expiratory carbon dioxide partial pressure sensor and pressure sensor in the breathing tubing 2. These signals are then transmitted to the signal processing and status analysis module. The microprocessor on the main control circuit board of the host 3 determines safety thresholds. When a safety threshold condition is triggered, the execution and output layer outputs according to the system's safety presets. When the user's physiological signals are within the normal range, the signal processing and status analysis module transmits the signals to the individualized impedance and breathing pattern training module. The microprocessor on the main control circuit board of the host 3 optimizes the dynamic load and breathing pattern. Subsequently, the scheme configuration module generates an individualized training prescription and transmits it to the main control module. The microprocessor on the main control circuit board of the host 3 performs scheduling and communication and transmits instructions to the execution and output layer. The adjustable impedance valve uses a stepper motor to control the rotary valve for precise resistance regulation, and the oxygen supply structure uses an oxygen flow regulating valve to control the oxygen supply amount.

[0102] Example 2

[0103] like Figures 2-4 As shown, this embodiment provides a ventilation-oxygenation dynamic balance-guided respiratory rehabilitation training control method, using the respiratory rehabilitation training system in Embodiment 1, including the following steps:

[0104] Step S1: Initialization and Signal Acquisition: Training begins. The user-signal input layer's human-computer interaction module loads the initialization training prescription. The user-signal input layer's multi-parameter physiological monitoring module, relying on the breathing tubing 2 and pulse oximetry sensor 4, acquires pressure signal P and blood oxygen saturation. and end-tidal carbon dioxide partial pressure ;

[0105] Step S2: Signal Processing: The signal processing and state analysis module of the intelligent control and decision-making layer calculates the flow rate Q and tidal volume in real time based on the pressure signal P. Inhalation time and exhalation time Characteristic parameters;

[0106] Step S3: Safety Threshold Judgment: Pre-set safety thresholds for various physiological parameters, compare monitoring data with threshold standards in real time, and immediately initiate a safety intervention sub-process when a parameter deviates from the safe range: First, determine whether blood oxygen-related parameters are abnormal, then classify and handle them based on the end-tidal carbon dioxide partial pressure value. According to different abnormal combinations, corresponding intervention measures such as stopping resistance training, adjusting oxygen supply concentration, and guiding standardized breathing methods are taken to ensure that the training process is safe and controllable; specifically, set safety thresholds, such as: or or or First, determine whether it occurs. or or or If so, immediately proceed to the security intervention sub-process: first determine if... or If not, then judge still ,like If resistance training is stopped, the oxygen concentration inhaled should be increased, and breathing should be deepened. Then stop resistance training and guide the user to exhale through pursed lips to reduce [the risk of infection]. Excretion; if it occurs or ,judge still ,like If resistance training is stopped, the oxygen concentration inhaled should be increased, and breathing should be deepened. Then stop resistance training, increase the inhaled oxygen concentration, guide the user to exhale with pursed lips, and reduce [the risk of infection]. discharge;

[0107] Step S4: When all physiological parameters are within the safe range, the individualized impedance and breathing pattern training module of the intelligent control and decision-making layer executes the individualized impedance and breathing pattern optimization algorithm, which specifically includes the following steps:

[0108] Step S401: Model Establishment: Establish a pre-model of respiratory pattern and ventilation-oxygenation demand based on the respiratory system motion equation, alveolar gas equation, and blood oxygenation dissociation curve equation;

[0109] The equation of motion for the respiratory system is expressed by the following formula:

[0110]

[0111]

[0112] In the formula, Indicates oral pressure; Indicates the force exerted by the respiratory muscles; Indicates the elasticity of the respiratory system; Indicates airway resistance; Indicates tidal volume; This indicates the resistance generated by the respiratory muscle trainer; Indicates positive end-expiratory pressure; Indicates alveolar ventilation; Indicates tidal volume; Indicates the volume of dead space;

[0113] It can be expressed by the following formula:

[0114]

[0115] In the formula, Indicates the amplitude of inspiratory muscle exertion; The time constant representing the exponential expiratory profile. ; Indicates the inhalation time; Indicates the exhalation time; Indicates the respiratory cycle;

[0116] The alveolar gas equation is expressed by the following formula:

[0117]

[0118] In the formula, Indicates alveolar oxygen partial pressure; Indicates the partial pressure of carbon dioxide in the alveoli; Indicates the fraction of inhaled oxygen concentration; Indicates atmospheric pressure; Indicates the partial pressure of water vapor; Indicates the respiratory exchange ratio;

[0119] The equation for the blood oxygenation dissociation curve is expressed by the following formula:

[0120]

[0121]

[0122]

[0123]

[0124]

[0125]

[0126]

[0127]

[0128] In the formula, Indicates blood oxygen saturation; Represents the Hill coefficient; This represents the partial pressure of oxygen when blood oxygen saturation is 50%. , , This represents the model's intrinsic parameters, obtained by fitting data from normal blood under standard physiological conditions. =2.82, =1.20, =29.25 mmHg; Indicates blood concentration; This indicates the concentration of hydrogen ions in the blood; This represents the standard half-saturated oxygen partial pressure, with a value of 26.8 mmHg; This represents the standard pH value, which is 7.4. This represents the standard CO2 partial pressure, with a value of 40 mmHg. It represents 2,3-diphosphoglyceric acid; This indicates the standard 2,3-DPG concentration, with values ​​ranging from [value missing]. M (mol / L); Indicates blood temperature; This represents the standard blood temperature, and its value is... ;

[0129] Step S402: Parameter Identification and Calculation: The user breathes at a calm state for a period of time. Using the user's oral pressure, flow rate, and tidal volume data, the respiratory system resistance R and elasticity E are identified. Specifically, the Levenberg-Marquardt algorithm is used to... Fit the model to make it conform to individual characteristics;

[0130] In the given discrete time points Data set obtained from the above measurement Estimate parameter vector ;

[0131] The objective function can be expressed as:

[0132]

[0133] In the formula, Indicates the first The residual of each data point is the difference between the measured stress and the model-predicted stress.

[0134] The residual can be defined as:

[0135] ;

[0136] Define residual vector Then the Jacobian matrix can be expressed as:

[0137]

[0138] The LM iterative equation can be expressed as: ,in ;

[0139] Step S403: Constructing the MPC-BO optimization problem: An individualized impedance and breathing pattern optimization algorithm, MPC-BO, is proposed based on the Model Predictive Control (MPC) framework and using Bayesian Optimization (BO) as the solver. The core idea of ​​this algorithm is to continuously optimize the equipment control parameters for the next training session based on the respiratory-gas exchange prediction model constructed according to the physiological mechanism after each breathing training session, transforming the clinical rehabilitation goal into a finite-time domain optimization control problem, and using BO to find the solution to this nonlinear, non-convex optimization problem.

[0140] The general form of the nonlinear MPC state-space equation of the system is:

[0141]

[0142] Wherein, the state vector Control input vector Output vector With interference vector They are represented as follows:

[0143] , , , ;

[0144] In the formula, Indicates the rate of carbon dioxide production;

[0145] Equations of state Represented as :

[0146]

[0147] In the formula, Represents the standard conversion constant;

[0148] Output equation Represented as :

[0149]

[0150] The forward Euler method is used to sample the continuous model at a fixed sampling period. Discretization yields the discrete-time prediction model:

[0151]

[0152] At the decision point after each training session, i.e. after the Tth training session, MPC is based on the current state. Solve a finite-time open-loop optimization problem to obtain the optimal control command for the next training iteration, i.e., the optimal control command for the (T+1)th training iteration.

[0153] The optimization objective is defined as minimizing the deviation between the physiological indicators and the target values ​​within the prediction time domain, while also considering control smoothness. Considering the periodic independence of respiratory muscle training (i.e., each cycle resets the state variables at the end of exhalation), both the prediction and control time domains are set to 1. The cost function is designed as follows:

[0154]

[0155] In the formula, , These represent the upper and lower limits of the clinically safe range for blood oxygen saturation, respectively. , These represent the upper and lower limits of the clinically safe range for the partial pressure of carbon dioxide at the end of expiration, respectively. , These represent the predicted values ​​of blood oxygen saturation and carbon dioxide partial pressure at the end of the respiratory tract, respectively. , , This represents the weighting coefficient, used to measure the importance of different objectives; This indicates the control amount from the previous cycle. The purpose is to prevent drastic changes in the control amount, such as a sudden increase in resistance from 30 cmH2O to 50 cmH2O, to avoid user discomfort.

[0156] BO solver in control variables Several sets are randomly generated within the feasible region, and the corresponding costs are obtained through the cost function. and form a sample point dataset. The distribution of the objective function is learned from the evaluated sample point dataset using a Gaussian process as a probabilistic surrogate model; the unknown cost function is assumed. Follows a Gaussian process:

[0157]

[0158] In the formula, This represents the mean function, and is usually set to a constant or zero; This represents the covariance function, also known as the kernel function, used to measure the input points. and Similarity between them;

[0159] The Matrn 5 / 2 kernel function was selected because it can effectively smooth the physiological response surface curve.

[0160]

[0161] In the formula, This represents the Euclidean distance between two points in the input space. It represents the signal variance, indicating the overall fluctuation range of the control output value; The length dimension represents the rate at which the correlation between input features decays.

[0162] In obtaining the observation dataset Then, for a new candidate point Its function value The posterior distribution is still a Gaussian distribution:

[0163] posterior mean It is the best estimate of the function value, while the posterior variance is... The uncertainty of this estimate has been quantified; the posterior mean and variance can be calculated using the following formulas:

[0164]

[0165]

[0166] In the formula, This represents the covariance matrix of the observed points; This represents the covariance vector between the new observation point and the previously observed points; Represents the observed cost vector; Indicates the variance of observation noise;

[0167] BO needs to maximize the acquisition function To determine the next assessment point This function balances the balance between Small area search, that is, developing the currently known optimal point, in Large-scale regional search relationships, i.e., exploring unknown areas to reduce uncertainty;

[0168] Using the expected improvement as the acquisition function, the current optimal observation value is defined as... The desired improved function is:

[0169]

[0170] In the formula, ; and Let represent the cumulative distribution function and probability density function of the standard normal distribution, respectively;

[0171] Step S404: Call the model: with This is the initial state. Simulate control commands to calculate the actual cost. And update the dataset ;

[0172] Step S405: Cost Evaluation: Select the control parameter combination with the lowest cost and output the optimal control command. ;

[0173] Step S5: Execution and Output: The main control module of the intelligent control and decision-making layer transmits the optimized instructions to the adjustable impedance execution module and oxygen supply module of the execution and output layer, which are then converted into execution operations by the stepper motor, oxygen supply flow regulating valve, and voice prompts.

[0174] Step S6: Loop and End: Repeat steps S1-S5 to perform signal acquisition, processing, judgment and individualized training parameter optimization for the next cycle until the training duration is reached, then end the training and the smart terminal 5 generates a training report.

[0175] It is understood that the present invention has been described through some embodiments, and those skilled in the art will recognize that various changes or equivalent substitutions can be made to these features and embodiments without departing from the spirit and scope of the invention. Furthermore, under the teachings of the present invention, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of the invention. Therefore, the present invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the protection scope of the present invention.

Claims

1. A respiratory rehabilitation training system guided by dynamic ventilation-oxygenation balance, characterized in that: It includes a user and signal input layer, an intelligent control and decision-making layer, and an execution and output layer. The user and signal input layer includes a multi-parameter physiological monitoring module, a human-computer interaction module, and a scheme configuration module. The supporting hardware components of the user and signal input layer include a breathing mouthpiece (1), a breathing tubing (2), a pulse oximetry sensor (4), and an intelligent terminal (5). The intelligent control and decision-making layer includes a signal processing and state analysis module, an individualized impedance and breathing pattern training module, and a main control module. The execution and output layer includes an adjustable impedance execution module and an oxygen supply module. The supporting hardware components of the intelligent control and decision-making layer and the execution and output layer are the host (3).

2. The ventilation-oxygenation dynamic balance-guided respiratory rehabilitation training system according to claim 1, characterized in that: The host (3) includes a main control circuit board with a microprocessor, an adjustable impedance valve, an oxygen supply structure, a power supply module and a status display screen. The signal processing and status analysis module of the intelligent control and decision layer, the individualized impedance and breathing mode training module and the main control module are all integrated on the microprocessor of the main control circuit board. The adjustable impedance execution module and the oxygen supply module of the execution and output layer are respectively integrated on the adjustable impedance valve and the oxygen supply structure.

3. The ventilation-oxygenation dynamic balance-guided respiratory rehabilitation training system according to claim 2, characterized in that: The adjustable impedance valve is driven by a stepper motor to rotate the valve and provides precise training resistance by changing the orifice diameter of the airflow channel. The oxygen supply structure is controlled by the oxygen supply flow regulating valve to control the oxygen supply. The breathing mouthpiece (1) is connected to the internal airway interface of the host (3) through the breathing tubing (2). The breathing tubing (2) integrates a pressure sensor and an end-expiratory carbon dioxide partial pressure sensor to collect the user's breathing pressure, breathing flow, tidal volume and end-expiratory carbon dioxide partial pressure. The pulse oximetry sensor (4) adopts any one of the finger clip type, patch type and wrist type and is connected to the signal acquisition interface of the host (3) through wired or wireless means to monitor the user's blood oxygen related physiological parameters in real time. The smart terminal (5) is a tablet computer or mobile phone, which establishes a data connection with the main control circuit board of the host (3) through Wi-Fi or Bluetooth to receive data, display real-time curves, provide training guidance and alarms, and complete the training program configuration.

4. A method for controlling respiratory rehabilitation training guided by dynamic ventilation-oxygenation balance, characterized in that: Using the respiratory rehabilitation training system according to any one of claims 1-3 includes the following steps: Step S1: Initialization and Signal Acquisition: Training begins. The user and signal input layer's human-computer interaction module loads the initialization training prescription. The user and signal input layer's multi-parameter physiological monitoring module, relying on the breathing tubing (2) and pulse oximetry sensor (4), acquires pressure signal P and blood oxygen saturation. and end-tidal carbon dioxide partial pressure ; Step S2: Signal Processing: The signal processing and state analysis module of the intelligent control and decision-making layer calculates the flow rate Q and tidal volume in real time based on the pressure signal P. Inhalation time and exhalation time ; Step S3: Safety threshold judgment: Pre-set the safety judgment thresholds for each physiological parameter, compare the monitoring data with the threshold standards in real time, and immediately start the safety intervention sub-process when the parameter deviates from the safe range: First, determine whether the blood oxygen related parameters are abnormal, and then classify and deal with them in combination with the end-tidal carbon dioxide partial pressure value. According to different abnormal combinations, take corresponding measures such as stopping resistance training, adjusting oxygen supply concentration, and guiding standardized breathing mode to ensure that the training process is safe and controllable. Step S4: When all physiological parameters are within the safe range, the individualized impedance and breathing pattern training module of the intelligent control and decision-making layer executes the individualized impedance and breathing pattern optimization algorithm. Step S5: Execution and Output: The main control module of the intelligent control and decision-making layer transmits the optimized instructions to the adjustable impedance execution module and oxygen supply module of the execution and output layer, which are then converted into execution operations by the stepper motor, oxygen supply flow regulating valve, and voice prompts. Step S6: Cycle and End: Repeat steps S1-S5 to collect, process, judge and optimize individualized training parameters for the next cycle until the training time is reached, end the training, and the smart terminal (5) generates a training report.

5. The method for controlling respiratory rehabilitation training guided by dynamic ventilation-oxygenation balance according to claim 4, characterized in that: In step S4, the individualized impedance and breathing pattern training module of the intelligent control and decision-making layer executes the individualized impedance and breathing pattern optimization algorithm, including the following steps: Step S401: Model building; Step S402: Parameter identification and calculation; Step S403: Construct the MPC-BO optimization problem; Step S404: Call the model; Step S405: Evaluate the cost.

6. The method for controlling respiratory rehabilitation training guided by dynamic ventilation-oxygenation balance according to claim 5, characterized in that: In step S401, a pre-model of respiratory mode and ventilation-oxygenation demand is established based on the respiratory system motion equation, alveolar gas equation, and blood oxygenation dissociation curve equation. The equation of motion for the respiratory system is expressed by the following formula: In the formula, Indicates oral pressure; Indicates the force exerted by the respiratory muscles; Indicates the elasticity of the respiratory system; Indicates airway resistance; Indicates tidal volume; This indicates the resistance generated by the respiratory muscle trainer; Indicates positive end-expiratory pressure; Indicates alveolar ventilation; Indicates tidal volume; Indicates the volume of dead space; It can be expressed by the following formula: In the formula, Indicates the amplitude of inspiratory muscle exertion; The time constant representing the exponential expiratory profile. ; Indicates the inhalation time; Indicates the exhalation time; Indicates the respiratory cycle; The alveolar gas equation is expressed by the following formula: In the formula, Indicates alveolar oxygen partial pressure; Indicates the partial pressure of carbon dioxide in the alveoli; Indicates the fraction of inhaled oxygen concentration; Indicates atmospheric pressure; Indicates the partial pressure of water vapor; Indicates the respiratory exchange ratio; The equation for the blood oxygenation dissociation curve is expressed by the following formula: In the formula, Indicates blood oxygen saturation; Represents the Hill coefficient; This represents the partial pressure of oxygen when blood oxygen saturation is 50%. , , This represents the model's intrinsic parameters, obtained by fitting data from normal blood under standard physiological conditions. =2.82, =1.20, =29.25 mmHg; Indicates blood concentration; This indicates the concentration of hydrogen ions in the blood. This represents the standard half-saturated oxygen partial pressure, with a value of 26.8 mmHg; This represents the standard pH value, which is 7.

4. Representation Standard The partial pressure is taken as 40 mmHg; It represents 2,3-diphosphoglyceric acid; This indicates the standard 2,3-DPG concentration, with values ​​ranging from [value missing]. M (mol / L); Indicates blood temperature; This represents the standard blood temperature, and its value is... .

7. The method for controlling respiratory rehabilitation training guided by dynamic ventilation-oxygenation balance according to claim 6, characterized in that: In step S402, the user is allowed to breathe calmly for a period of time. The respiratory system resistance R and elasticity E are identified using data on oral pressure, flow rate, and tidal volume. The Levenberg-Marquardt algorithm is then used to... Fit the model to make it conform to individual characteristics; In the given discrete time points Data set obtained from the above measurement Estimate parameter vector ; The objective function can be expressed as: In the formula, Indicates the first The residual of each data point is the difference between the measured stress and the model-predicted stress. The residual can be defined as: ； Define residual vector Then the Jacobian matrix can be expressed as: The LM iterative equation can be expressed as: ,in .

8. The method for controlling respiratory rehabilitation training guided by dynamic balance of ventilation and oxygenation according to claim 7, characterized in that: In step S403, an individualized impedance and breathing pattern optimization algorithm MPC-BO is proposed, which is based on the Model Predictive Control (MPC) framework and uses Bayesian Optimization (BO) as the solver. After each breathing training session, the algorithm continuously optimizes the equipment control parameters for the next training session based on the respiratory-gas exchange prediction model constructed from physiological mechanisms. This transforms the clinical rehabilitation goal into a finite-time domain optimization control problem, and uses BO to find the solution to this nonlinear, non-convex optimization problem. The general form of the nonlinear MPC state-space equation of the system is: Wherein, the state vector Control input vector Output vector With interference vector They are represented as follows: ， ， ， ； In the formula, Indicates the rate of carbon dioxide production; Equations of state Represented as : In the formula, Represents the standard conversion constant; Output equation Represented as : The forward Euler method is used to sample the continuous model at a fixed sampling period. Discretization yields the discrete-time prediction model: At the decision point after each training session, i.e. after the Tth training session, MPC is based on the current state. Solve a finite-time open-loop optimization problem to obtain the optimal control command for the next training iteration, i.e., the optimal control command for the (T+1)th training iteration. The optimization objective is defined as minimizing the deviation between the physiological indicators and the target values ​​within the prediction time domain, while also considering control smoothness. Considering the periodic independence of respiratory muscle training (i.e., each cycle resets the state variables at the end of exhalation), and setting both the prediction and control time domains to 1, the cost function is designed as follows: In the formula, , These represent the upper and lower limits of the clinically safe range for blood oxygen saturation, respectively. , These represent the upper and lower limits of the clinically safe range for the partial pressure of carbon dioxide at the end of expiration, respectively. , These represent the predicted values ​​of blood oxygen saturation and carbon dioxide partial pressure at the end of the respiratory tract, respectively. , , This represents the weighting coefficient, used to measure the importance of different objectives; This indicates the control value from the previous cycle, preventing drastic changes in the control value and avoiding user discomfort. BO solver in control variables Several sets are randomly generated within the feasible region, and the corresponding costs are obtained through the cost function. and form a sample point dataset. The distribution of the objective function is learned from the evaluated sample point dataset using a Gaussian process as a probabilistic surrogate model; the unknown cost function is assumed. Follows a Gaussian process: In the formula, This represents the mean function, and is usually set to a constant or zero; This represents the covariance function, also known as the kernel function, used to measure the input points. and Similarity between them; The Matrn 5 / 2 kernel function is selected because it can smooth the physiological response surface curve. In the formula, This represents the Euclidean distance between two points in the input space. It represents the signal variance, indicating the overall fluctuation range of the control output value; The length dimension represents the rate at which the correlation between input features decays. In obtaining the observation dataset Then, for a new candidate point Its function value The posterior distribution is still a Gaussian distribution: Posterior mean It is the best estimate of the function value, posterior variance. Quantify the uncertainty of this estimate; the posterior mean and variance can be calculated using the following formulas: In the formula, This represents the covariance matrix of the observed points; This represents the covariance vector between the new observation point and the previously observed points; Represents the observed cost vector; Indicates the variance of observation noise; BO Maximization Acquisition Function To determine the next assessment point This function balances the balance between Small area search, that is, developing the currently known optimal point, in Large-scale regional search relationships, i.e., exploring unknown areas to reduce uncertainty; Using the expected improvement as the acquisition function, the current optimal observation value is defined as... The desired improved function is: In the formula, ; and Let represent the cumulative distribution function and probability density function of the standard normal distribution, respectively.

9. The method for controlling respiratory rehabilitation training guided by dynamic balance of ventilation and oxygenation according to claim 8, characterized in that: In step S404, with This is the initial state. Simulate control commands to calculate the actual cost. And update the dataset .

10. The method for controlling respiratory rehabilitation training guided by dynamic ventilation-oxygenation balance according to claim 9, characterized in that: In step S405, the control parameter combination with the lowest cost is selected, and the optimal control command is output. .

Images