AI-based respiratory exercise device and system for patients with chronic obstructive pulmonary disease

Abstract

The present application relates to the technical field of respiratory training apparatus, in particular to a respiratory exercise device and system for chronic obstructive pulmonary disease patients based on AI, wherein the device comprises a blowing nozzle and an AI control system; one end of the blowing nozzle is fixedly connected with an elbow, the distal end of the elbow away from the blowing nozzle is fixedly connected with a test airway, a test ball is slidably arranged in the test airway, the distal end of the test airway away from the elbow is provided with an adjusting assembly for adjusting the air output of the test airway, and a driving assembly for driving the adjusting assembly to operate; a monitoring assembly for monitoring the gas flow value in the test airway is arranged in the test airway; the driving assembly is used for driving the adjusting assembly to adjust the air output of the test airway based on the flow value. The present application can realize dynamic self-adaptive adjustment of respiratory resistance, and improve the accuracy and safety of lung rehabilitation training for COPD patients.

Description

Technical Field

[0001] This invention relates to the field of respiratory training equipment technology, specifically to an AI-based respiratory training device and system for patients with chronic obstructive pulmonary disease. Background Technology

[0002] Chronic obstructive pulmonary disease (COPD) is a common, preventable, and treatable chronic respiratory disease characterized by persistent respiratory symptoms and irreversible airflow limitation, with high global prevalence and mortality rates. Clinical studies have confirmed that both inspiratory and expiratory muscle training can effectively improve respiratory muscle strength, exercise capacity, and quality of life in COPD patients, and the combined training of the two shows synergistic advantages in improving dyspnea and exercise tolerance.

[0003] Currently, the main respiratory training devices on the market are resistance load / threshold load respiratory trainers. Typical products include the Threshold® series of devices, which require patients to generate sufficient pressure during breathing to open the airway through preset springs or magnetically controlled threshold valves, thereby applying load training to the respiratory muscles.

[0004] However, in actual use of the aforementioned breathing training device, which employs fixed resistance levels or mechanically preset thresholds, patients need to manually adjust the resistance level. Because the respiratory muscle endurance of COPD patients changes non-linearly with training duration and fatigue levels, fixed resistance is difficult to dynamically adjust based on the peak flow rate, respiratory rate, and waveform characteristics of each breath. This leads to excessive resistance in the early stages of training, easily inducing respiratory muscle over-fatigue, and insufficient resistance in the later stages, failing to provide effective load stimulation, thus limiting the effectiveness of pulmonary rehabilitation training to some extent. Therefore, it is necessary to propose an AI-based breathing training device and system for patients with chronic obstructive pulmonary disease to address these issues. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides an AI-based respiratory training device and system for patients with chronic obstructive pulmonary disease (COPD), which enables dynamic adaptive adjustment of respiratory resistance, thereby improving the accuracy and safety of pulmonary rehabilitation training for COPD patients.

[0006] To achieve the above objectives, the technical solution of the present invention is as follows: an AI-based respiratory exercise device for patients with chronic obstructive pulmonary disease, comprising an air nozzle and an AI control system; one end of the air nozzle is fixedly connected to a bend, and the end of the bend away from the air nozzle is fixedly connected to a test airway, in which a test ball is slidably fitted; the end of the test airway away from the bend is provided with an adjustment component for adjusting the air output of the test airway, and a drive component for driving the adjustment component to operate; a monitoring component is provided in the test airway for monitoring the gas flow rate value in the test airway; the drive component is used to drive the adjustment component to adjust the air output of the test airway based on the flow rate value.

[0007] The technical principles of the above solution are as follows:

[0008] The user places the mouthpiece over their mouth and blows or inhales into it. The test ball is lifted or lowered by the airflow. The patient controls their breathing rhythm by observing the position and state of the test ball. The monitoring component within the test airway collects the instantaneous flow rate, peak flow rate, respiratory rate, and waveform characteristics of each breath. Using a pre-trained respiratory mechanics model, the system dynamically determines the patient's current respiratory muscle load adaptation stage. When the system determines that the existing resistance does not match the patient's real-time breathing capacity, the drive component immediately drives the adjustment component to change the effective airflow cross-sectional area at the distal end of the test airway. This enables the AI ​​control system to automatically adjust the resistance during inhalation or exhalation, forming a control mechanism of real-time monitoring, dynamic judgment, and automatic adjustment. This ensures that the training load always adapts to the patient's non-linearly changing respiratory muscle endurance.

[0009] The above approach has the following beneficial effects:

[0010] 1. This solution achieves dynamic adaptive adjustment of respiratory resistance. By collecting data such as peak flow rate and waveform characteristics in real time through monitoring components, and combining with AI models to dynamically determine the patient's fatigue state, it effectively avoids the problems of excessive initial load and insufficient stimulation at the end caused by fixed resistance, thereby improving the safety and comfort of the training process.

[0011] 2. This solution uses a mechanical structure of test airway, test ball, adjustment component and drive component combined with AI control, which can automatically and smoothly adjust the effective air output cross-sectional area within a single respiratory cycle. This eliminates the need for patients to manually change gears or adjust valves, reducing the operational threshold and improving compliance with long-term pulmonary rehabilitation training.

[0012] 3. This program monitors the mechanical characteristics of each breath in real time, ensuring that the training load always matches the patient's non-linearly changing respiratory muscle endurance. This prevents respiratory muscle damage caused by over-fatigue and ensures that each breath generates effective load stimulation, thereby improving the overall rehabilitation effect of combined training of inspiratory and expiratory muscles.

[0013] Furthermore, the adjustment assembly includes an air outlet with a hexagonal groove on its side wall; a top plate rotatably fitted onto the air outlet, with several guide grooves circumferentially formed on the top plate; several triangular plates are provided between the top plate and the air outlet, each triangular plate having a slider fixedly connected to it, the sliders slidingly engaging with the grooves, and each triangular plate also having a sliding rod fixedly connected to it, the sliding rods slidingly engaging with their adjacent guide grooves; and a ring gear is fitted over the top plate.

[0014] Beneficial effects: Through the cooperative structure of hexagonal grooves, guide grooves, and triangular plate sliders, the rotational motion of the ring gear is converted into the synchronous radial expansion and contraction of multiple triangular plates, thereby continuously and smoothly changing the effective ventilation cross-sectional area of ​​the air outlet. The hexagonal layout ensures the symmetry and consistency of the movement of the multi-lobed triangular plates, making the airflow resistance adjustment more uniform and reliable.

[0015] Furthermore, the drive assembly includes an electronically controlled cylinder fixedly connected to the side wall of the air outlet, an AI control system for controlling the extension and retraction of the output shaft of the electronically controlled cylinder, and a rack fixedly connected to the output shaft of the electronically controlled cylinder, the rack meshing with a ring gear.

[0016] Beneficial effects: By using an electric cylinder in conjunction with a rack and pinion gear transmission, the AI ​​control system can achieve angle control of the exhaust cross-sectional area.

[0017] Furthermore, the monitoring components include a gas flow sensor fixedly connected to the inner wall of the test airway, and an AI control system for receiving the gas flow value in the test airway sent by the gas flow sensor and controlling the extension and retraction of the output shaft of the electric cylinder based on the gas flow value; the test airway is equipped with an anti-choking component to prevent the test ball from entering the patient's airway.

[0018] Beneficial effects: By directly embedding the gas flow sensor into the test airway, it is possible to acquire the instantaneous flow rate, peak flow rate, respiratory waveform and frequency data of each breath of the patient at a millisecond sampling rate, so that the resistance adjustment is truly based on the patient's real-time respiratory mechanics characteristics, realizing closed-loop intelligent training with simultaneous measurement and adjustment.

[0019] Furthermore, the anti-choking component includes limiting blocks that are symmetrically fixedly connected to the inner wall of the test airway.

[0020] Beneficial effects: The limiting blocks are symmetrically set on the inner wall of the test airway, which can effectively limit the maximum movement of the test ball during inhalation or exhalation, and prevent the patient from inhaling the test ball due to excessive force or sudden coughing.

[0021] Furthermore, the test showed that handles were symmetrically fixed to the sidewalls of the airway.

[0022] Beneficial effects: The grip provides patients with a stable handhold, while the symmetrical design facilitates alternating use with the left and right hands, adapting to the operating habits of different patients.

[0023] Furthermore, AI-based respiratory training systems for patients with chronic obstructive pulmonary disease include:

[0024] The data acquisition module is used to collect exercise data during the patient's breathing exercises using a gas flow sensor and then send the exercise data to the AI ​​analysis module.

[0025] The AI ​​analysis module is used to input respiratory exercise sample data from patients with different COPD severity levels into a CNN neural network model for learning and training, thereby obtaining a respiratory function assessment model for COPD patients. The AI ​​analysis module is also used to input exercise data into the respiratory function assessment model for COPD patients, generate personalized exercise plans, and send the personalized exercise plans to the control module.

[0026] The control module is used to convert personalized exercise programs into control commands and send control commands to the drive components. It also uses these control commands to control the extension and retraction of the electronically controlled cylinder to adjust the air output of the test airway.

[0027] Beneficial effects: By learning from respiratory exercise sample data of patients with different COPD severity levels using a CNN neural network model, it can automatically extract the complex mapping relationship between multidimensional features such as respiratory flow, frequency, and duration and airway resistance, thereby generating highly personalized exercise programs.

[0028] Furthermore, the training data in the acquisition module includes gas flow rate and respiratory rate data.

[0029] Beneficial effects: Simultaneous collection of gas flow rate and respiratory rate data allows for a more comprehensive characterization of the biomechanical features of each breath. Flow rate data reflects the instantaneous exertion capacity of the respiratory muscles, while respiratory rate data reflects the endurance and fatigue level of the respiratory muscles. Combining these two data points provides more complete input features for the AI ​​analysis module, thereby improving the accuracy of the respiratory function assessment model and the scientific validity of personalized training programs.

[0030] Furthermore, in the AI ​​analysis module, the breathing exercise sample data includes respiratory flow, respiratory rate, respiratory duration, airway resistance adaptation parameters, and corresponding respiratory function assessment results.

[0031] Beneficial effects: By using respiratory flow rate, respiratory rate, respiratory duration, airway resistance adaptation parameters, and corresponding respiratory function assessment results as training samples, the neural network model can learn the nonlinear relationship between patients' respiratory performance and functional improvement under different resistance loads. This multi-dimensional sample design enhances the model's generalization ability, enabling it to recommend optimal resistance change curves for COPD patients with different disease stages, achieving truly personalized intelligent respiratory training.

[0032] Furthermore, the AI ​​analysis module includes a feedback optimization subunit, which is used to send the data from each breathing exercise as new sample data to the COPD patient respiratory function assessment model for incremental learning after each patient completes the exercise.

[0033] Beneficial effects: By introducing feedback optimization sub-units, the COPD patient respiratory function assessment model possesses continuous learning and adaptive evolution capabilities. This closed-loop mechanism effectively overcomes the limitation of static models in tracking individual rehabilitation trajectories. With increased usage, the personalized exercise programs generated by the model will increasingly match the patient's current actual respiratory capacity, further improving the accuracy of pulmonary rehabilitation training and the long-term intervention effect. Attached Figure Description

[0034] Figure 1 This is an isometric view of the AI-based breathing exercise device for patients with chronic obstructive pulmonary disease according to the present invention.

[0035] Figure 2 This is a lateral cross-sectional view of the test airway in the AI-based respiratory training device for patients with chronic obstructive pulmonary disease according to the present invention.

[0036] Figure 3 for Figure 2 Sectional view along the AA direction.

[0037] Figure 4 for Figure 2 Sectional view along the BB direction.

[0038] Figure 5 for Figure 2 A cross-sectional view along the CC direction.

[0039] Figure 6 This is a system structure diagram of the AI-based respiratory exercise system for patients with chronic obstructive pulmonary disease according to the present invention.

[0040] The reference numerals in the accompanying drawings of the instruction manual include: 1. Air nozzle; 2. Elbow; 3. Test air passage; 4. Test ball; 5. Air outlet; 6. Slide groove; 7. Top plate; 8. Guide groove; 9. Triangular plate; 10. Slider; 11. Slide rod; 12. Ring gear; 13. Electric cylinder; 14. Rack; 15. Limiting block; 16. Handle; 17. Housing. Detailed Implementation

[0041] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0042] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0043] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0044] The following detailed description illustrates the specific implementation method:

[0045] Implementation, for example, attached Figure 1 As shown:

[0046] This AI-based breathing exercise device for patients with chronic obstructive pulmonary disease includes a mouthpiece 1 and an AI control system. One end of the mouthpiece 1 is integrally formed and connected to a bend 2. The end of the bend 2 away from the mouthpiece 1 is connected to a test airway 3, and a test ball 4 slides within the test airway 3. The bend 2 creates an airflow deflection between the mouthpiece 1 and the test airway 3, preventing saliva from directly splashing into the test airway 3 when the patient blows air. It also brings the overall center of gravity of the device closer to the hand, making it easier to operate.

[0047] like Figure 2 As shown, in this embodiment, the test airway 3 is a transparent rigid plastic round tube, and the test ball 4 is a hollow sphere made of polypropylene. The outer diameter of the test ball 4 is 2mm smaller than the inner diameter of the test airway 3, so that the test ball 4 can be pushed by the airflow to slide freely along the axial direction of the test airway 3, and at the same time, the airflow can pass through the gap between the test ball 4 and the test airway 3.

[0048] The end of the test air passage 3 away from the bend 2 is provided with an adjustment component for adjusting the air output of the test air passage 3, and a drive component for driving the adjustment component to operate.

[0049] The test airway 3 is equipped with a monitoring component for monitoring the gas flow rate within the test airway 3; the drive component is used to drive the adjustment component to adjust the gas output of the test airway 3 based on the flow rate. Handles 16 are symmetrically fixed to the side wall of the test airway 3 by screws.

[0050] like Figure 2 and Figure 5 As shown, specifically, the adjustment component includes an air outlet 5, and a hexagonal groove 6 is opened on the side wall of the air outlet 5.

[0051] like Figure 3 , Figure 4 and Figure 5 As shown, a top plate 7 is rotatably fitted onto the air outlet 5. The top plate 7 has several guide grooves 8 circumferentially open. Several triangular plates 9 are provided between the top plate 7 and the air outlet 5. Each triangular plate 9 has an integrally formed slider 10, which slides in a groove 6. Each triangular plate 9 also has an integrally formed sliding rod 11, which slides in a groove 8 adjacent to it. A ring gear 12 is integrally formed and fitted onto the top plate 7. In this embodiment, a housing 17 is fixedly connected to the air outlet 5 by screws. The housing 17 protects the adjustment components on the air outlet 5, thus preventing damage.

[0052] like Figure 3 As shown, specifically, the drive assembly includes an electric cylinder 13 fixedly connected to the side wall of the air outlet 5 by screws. The AI ​​control system is used to control the extension and retraction of the output shaft of the electric cylinder 13. A rack 14 is fixedly connected to the output shaft of the electric cylinder 13 by screws, and the rack 14 meshes with the ring gear 12.

[0053] Combination Figure 1 and Figure 2 As shown, the user points the mouthpiece 1 to their mouth and blows or inhales into it. The test ball 4 is lifted or lowered by the airflow. The patient controls their breathing rhythm by observing the position and state of the test ball 4. The monitoring component in the test airway 3 collects the instantaneous flow rate, peak flow rate, respiratory rate, and waveform characteristics of each breath. Using a pre-trained respiratory mechanics model, the system dynamically determines the current load adaptation stage of the patient's respiratory muscles. When it is determined that the existing resistance does not match the patient's real-time breathing capacity, the drive component immediately drives the adjustment component to change the effective air outlet cross-sectional area at the distal end of the test airway 3. This enables the AI ​​control system to automatically adjust the resistance of inhalation or exhalation, forming a control mechanism of real-time monitoring, dynamic judgment, and automatic adjustment. This ensures that the training load always adapts to the patient's nonlinearly changing respiratory muscle endurance.

[0054] Combination Figure 3 As shown, when the output shaft of the electric control cylinder 13 extends or retracts, the rack 14 drives the ring gear 12 to rotate, and the ring gear 12 drives the top plate 7 to rotate on the air outlet 5. The guide groove 8 on the top plate 7 pushes each triangular plate 9 through the slide rod 11, so that each triangular plate 9 shrinks inward or expands outward in a radial direction under the constraint of the hexagonal slide groove 6.

[0055] When the triangular plate 9 contracts radially in sync, the effective ventilation cross-sectional area in the center of the air outlet 5 decreases, and at this time, the airflow resistance increases.

[0056] When the triangular plate 9 expands radially in sync, the effective ventilation cross-sectional area increases, and the airflow resistance decreases. This adjustment process is continuous and smooth, and can be completed within a single respiratory cycle of the patient.

[0057] like Figure 2 As shown, specifically, the monitoring components include a gas flow sensor (not shown in the figure) that is fixedly connected to the inner wall of the test airway 3 by screws. The AI ​​control system is used to receive the gas flow value in the test airway 3 sent by the gas flow sensor and control the extension and retraction of the output shaft of the electric cylinder 13 based on the gas flow value. The test airway 3 is equipped with an anti-choking component to prevent the test ball 4 from entering the patient's airway.

[0058] like Figure 2 As shown, specifically, the anti-choking component includes a limiting block 15 symmetrically and integrally formed on the inner wall of the test airway 3.

[0059] Combination Figure 1 As shown, based on the patient's condition and individual circumstances, corresponding breathing exercise standards are formulated, and the flow rate threshold is set for the AI ​​control system according to the breathing exercise standards. Before use, the patient holds the handle 16 on the side wall of the test airway 3 with both hands, and puts the mouthpiece 1 against the lips. After the device is started, the gas flow sensor begins to monitor the gas flow rate in the test airway 3 in real time and transmits the monitoring data to the AI ​​control system in real time.

[0060] When the patient performs blowing exercises, the airflow enters the bend 2 through the blowing nozzle 1, turns after the bend 2 and flows into the test airway 3. The airflow pushes the test ball 4 in the test airway 3 to slide along the axis of the test airway 3 towards the outlet 5. The patient can visually observe the sliding distance of the test ball 4 through the transparent test airway 3 and judge his own blowing force. In this embodiment, the test ball 4 is made of colored material for easy observation by the user.

[0061] Simultaneously, the gas flow sensor transmits the real-time monitored airflow value to the AI ​​control system. The AI ​​control system compares this flow value with a preset threshold. If the flow value is lower than the preset lower limit, it indicates that the patient's blowing force is insufficient. The AI ​​control system controls the output shaft of the electric cylinder 13 to retract, driving the rack 14 to move, which in turn drives the ring gear 12 and the top plate 7 to rotate, causing the triangular plate 9 to expand radially synchronously, increasing the effective ventilation cross-sectional area of ​​the air outlet 5, reducing airflow resistance, and making it easier for the patient to increase the blowing flow to achieve the training goal. If the flow value is higher than the preset upper limit, it indicates that the patient's blowing force is too great. The AI ​​control system controls the output shaft of the electric cylinder 13 to extend, driving the triangular plate 9 to contract radially synchronously, reducing the effective ventilation cross-sectional area of ​​the air outlet 5, increasing airflow resistance, guiding the patient to adjust the blowing force, and avoiding excessive force that could damage the respiratory tract.

[0062] When the patient performs inhalation exercises, the limiting block 15 acts as a barrier to prevent the test ball 4 from entering the mouthpiece 1 from the elbow 2 and then into the patient's mouth, thus preventing the risk of suffocation. Throughout the breathing exercise, the adjustment mechanism moves smoothly and continuously, and can be completed within a single inhalation or exhalation cycle without requiring manual operation by the patient, reducing the difficulty of operation and making it suitable for elderly patients and patients with severe chronic obstructive pulmonary disease.

[0063] like Figure 6 As shown, an AI-based respiratory training system for patients with chronic obstructive pulmonary disease includes:

[0064] The data acquisition module uses a gas flow sensor to collect exercise data during the patient's breathing exercises and sends the data to the AI ​​analysis module. The exercise data includes gas flow rate and respiratory rate data.

[0065] The AI ​​analysis module is used to input respiratory exercise sample data from COPD patients at different disease levels into a CNN neural network model for learning and training, resulting in a COPD patient respiratory function assessment model. The AI ​​analysis module also inputs exercise data into the COPD patient respiratory function assessment model to generate personalized exercise plans, which are then sent to the control module. The respiratory exercise sample data includes respiratory flow, respiratory rate, respiratory duration, airway resistance adaptation parameters, and corresponding respiratory function assessment results.

[0066] The AI ​​analysis module includes a feedback optimization subunit, which sends the data from each breathing exercise as new sample data to the COPD patient respiratory function assessment model for incremental learning.

[0067] The control module is used to convert personalized exercise programs into control commands and send control commands to the drive components, and use the control commands to control the extension and retraction of the electronically controlled cylinder 13 to adjust the air output of the test airway 3.

[0068] Combination Figure 6 As shown, the COPD patient respiratory function assessment model in the AI ​​analysis module uses a CNN neural network. First, it filters the respiratory exercise sample data of COPD patients with different severity levels (mild, moderate, severe, and very severe). Then, it standardizes the filtered valid sample data. At the same time, it labels the respiratory function assessment results corresponding to each sample data, converting the assessment results (such as normal respiratory function, mild impairment, moderate impairment, and severe impairment) into numerical labels that the model can recognize, and establishing the correspondence between sample data and assessment results.

[0069] A CNN neural network model was constructed, consisting of 3 convolutional layers, 2 pooling layers, 2 fully connected layers, and 1 output layer. The convolutional layers used 3×3 kernels to extract key features from the sample data (such as peak respiratory flow and periodicity of respiratory rate), and the ReLU activation function was used to enhance the model's nonlinear fitting ability and avoid gradient vanishing. The pooling layers used max pooling with a 2×2 kernel to reduce the dimensionality of the features extracted by the convolutional layers while retaining key feature information. The fully connected layer integrated the feature vectors output by the pooling layers, mapping them to a predefined feature space to achieve deep feature fusion. The output layer used the Softmax activation function to output the probability distribution of the patient's respiratory function assessment, corresponding to different levels of respiratory function impairment, and finally outputting the assessment result with the highest probability and the corresponding personalized training program adaptation parameters.

[0070] During device operation, the acquisition module collects real-time gas flow and respiratory rate data during the patient's breathing exercise through a gas flow sensor. This data is then transmitted to the AI ​​analysis module and input into the trained COPD patient respiratory function assessment model. The COPD patient respiratory function assessment model extracts features from the real-time data through convolutional and pooling layers, integrates them through a fully connected layer, and outputs the patient's current respiratory function assessment results. Simultaneously, it combines airway resistance adaptation parameters from the sample data to generate a personalized exercise plan adapted to the patient's current respiratory function state (including preset flow thresholds, airway resistance adjustment gradients, and exercise duration suggestions). This personalized exercise plan is then sent to the control module to achieve AI adaptive adjustment during the exercise process.

[0071] In this embodiment, the AI ​​analysis module includes a patient information input unit, which is used to input the patient's age, height, weight, COPD classification (GOLD 1–4), and number of previous acute exacerbations. Based on the patient's classification, the AI ​​analysis module automatically generates initial upper flow rate thresholds (F_high) and lower flow rate thresholds (F_low), and sends these to the AI ​​control system via the control module. For example, in this embodiment: GOLD 2 patients: F_low = 15 L / min, F_high = 30 L / min; GOLD 3 patients: F_low = 10 L / min, F_high = 22 L / min.

[0072] The patient holds the handle 16 with both hands, places the mouthpiece 1 against their lips, and begins to exhale. The gas flow sensor monitors the flow rate F_actual in the test airway 3 in real time at a sampling frequency of ≥100Hz and transmits the data to the AI ​​control system.

[0073] If F_actual<F_low (insufficient blowing): the AI control system drives the output shaft of the electric control cylinder 13 to retract, at this time the rack 14 drives the ring gear 12 to rotate clockwise, the top plate 7 rotates clockwise, and each triangular plate 9 expands radially outward, at this time the effective ventilation cross-sectional area of the air outlet nozzle 5 increases, the airflow resistance decreases, and the patient can maintain or raise the height of the ball without more effort, and obtain positive encouragement.

[0074] If F_actual>F_high (violent blowing): the AI control system drives the output shaft of the electric control cylinder 13 to extend, the rack 14 drives the ring gear 12 to rotate counterclockwise, the triangular plate 9 contracts radially inward, and the effective ventilation cross-sectional area decreases. At this time, the airflow resistance increases sharply, the rise of the ball slows down or even slightly falls back, the patient immediately perceives "blowing too hard", thereby actively reducing the expiratory force, so that the patient's blowing gradually tends to be stable, and the exercise effect is improved.

[0075] If F_low≤F_actual≤F_high (normal range): the AI control system maintains the current position of the triangular plate 9 without adjustment.

[0076] In this embodiment, based on the patient's training cycle, the AI analysis module also dynamically adjusts the airway resistance by combining the exercise data accumulated by the acquisition module (including indicators such as flow compliance rate, average respiratory flow, and respiratory stability in each cycle) to achieve a gradual exercise effect. The specific adjustment method is as follows: for example, in this embodiment, the standard training cycle is set to 4 weeks, and each week is an adjustment cycle. The AI analysis module summarizes and analyzes the patient's exercise data every week to determine the recovery of the patient's respiratory function. For GOLD 1-2 grade mild to moderate patients, the initial cycle (the first week) maintains the initial resistance and flow threshold. If the patient's flow compliance rate ≥ 80% within the first week (that is, the proportion of time when F_low≤F_actual≤F_high in each exercise is ≥ 80%), and the respiratory frequency fluctuation ≤ 10%, then in the second week, the control module will lower F_high by 5% and raise F_low by 5%, while controlling the electric control cylinder 13 to fine-tune the position of the triangular plate 9, so that the effective ventilation cross-sectional area of the air outlet nozzle 5 decreases by 3%-5%, slightly increasing the airflow resistance; in the third week, if the patient can still maintain a compliance rate ≥ 80%, continue to adjust the resistance and flow threshold according to the above gradient; after the adjustment in the fourth week is completed, the AI analysis module generates a resistance adjustment benchmark for the next stage (the subsequent 4 weeks) by combining the patient's 4-week cumulative data.

[0077] For patients with GOLD grade 3-4 severe to very severe conditions, the adjustment cycle is extended to 2 weeks, and the resistance adjustment gradient is halved (the effective ventilation cross-sectional area is reduced by 1.5%-2.5% each time, F_high is reduced by 2.5%, and F_low is increased by 2.5%). The next adjustment can only be carried out if the target achievement rate is ≥70% for 2 consecutive weeks to avoid the resistance from increasing too quickly and aggravating the patient's airway burden. If the patient's target achievement rate is <60% in a certain adjustment cycle, the AI ​​analysis module will automatically revert to the resistance and flow thresholds of the previous cycle and extend the training time of that cycle until the patient's target achievement rate meets the requirements before continuing to advance the adjustment.

[0078] Throughout the entire cycle of adjustment, the AI ​​analysis module records the correlation data between resistance adjustment parameters and the patient's respiratory function in real time, continuously optimizes the adjustment gradient, and ensures that the resistance adjustment is adapted to the patient's recovery progress, so as to ensure the training effect and avoid damage caused by over-training.

[0079] In other embodiments, the AI-based respiratory exercise system for patients with chronic obstructive pulmonary disease also includes a patient-side application. After installing the application on a mobile device, patients can complete AI-powered intelligent management functions, specifically:

[0080] Disease screening and risk assessment: Before the patient uses the application for the first time, the patient-side application collects the patient's basic information through an interactive questionnaire, including age, gender, smoking history, number of previous acute exacerbations, current medication status, and CAT score. The AI ​​system quickly assesses the patient's disease status based on the built-in COPD preliminary screening model, generates a preliminary risk level (low risk, medium risk, high risk), and recommends the corresponding initial training intensity and frequency according to the risk level.

[0081] Medication reminders and adherence management: The AI ​​system automatically generates a personalized medication reminder plan based on the patient's entered medication regimen (including inhaler type, usage, dosage, and administration time). The patient-side application reminds patients to take their medication on time via push notifications, vibrations, etc., at preset times. After completing the medication, patients can check in and confirm in the application.

[0082] Rehabilitation check-in and real-time feedback on training status: Before each use of the breathing exercise device, patients need to complete a rehabilitation check-in in the patient-side application, recording their subjective feelings for the day (such as the degree of difficulty breathing, fatigue, coughing, and sputum production). During training, the device collects training data in real time through a gas flow sensor and dynamically displays the flow curve, test ball position simulation animation, and target achievement status in a visual chart format on the patient-side application interface. Patients can intuitively understand the performance of each breath and receive immediate positive encouragement (such as "This exhalation is up to standard, keep it up!") or guidance prompts (such as "The exhalation force is too weak, please use more force"), enhancing the fun and compliance of training.

[0083] Multimodal physiological data fusion assessment: In this embodiment, the breathing training device also integrates or can be wirelessly connected to a fingertip pulse oximeter. During training, the pulse oximeter collects the patient's blood oxygen saturation (SpO2) and heart rate data in real time and transmits them to the AI ​​analysis module via Bluetooth. The AI ​​analysis module performs multimodal fusion analysis on gas flow data, respiratory rate data, and blood oxygen and pulse data to comprehensively assess the patient's respiratory function status and training intensity adaptability. The specific assessment method is as follows:

[0084] When the blood oxygen saturation remains below 90% or the heart rate exceeds (220 - age) × 85% during training, the AI ​​analysis module determines that the current training intensity is too high. It immediately drives the component to increase the effective ventilation cross-sectional area of ​​the air outlet 5 of the test airway 3 through the control module command to reduce airflow resistance. At the same time, the patient's application pops up a warning message: "Blood oxygen is too low / heart rate is too fast. Please stop training and rest."

[0085] When blood oxygen saturation is maintained above 92%, heart rate is within a safe range, and flow rate meets the target at ≥70% during training, the AI ​​analysis module determines that the training intensity is appropriate and maintains the current resistance setting.

[0086] When the patient's average flow rate shows an upward trend and the heart rate recovery time gradually shortens during five consecutive training sessions, the AI ​​analysis module determines that the patient's respiratory function has improved and automatically adjusts the training difficulty appropriately in the next training cycle (e.g., reducing F_high by 3% and increasing F_low by 3%) to achieve a gradual and personalized pulmonary rehabilitation plan.

[0087] The acquisition module is also used to receive blood oxygen saturation data and heart rate data sent by the blood oxygen pulse sensor, and synchronously send the above multimodal data to the AI ​​analysis module. The AI ​​analysis module also includes a multimodal fusion assessment subunit, which is used to perform time alignment and feature fusion of gas flow data, respiratory rate data, blood oxygen saturation data, and heart rate data to construct a complete physiological state profile of the patient for each training session, thereby more accurately judging whether the training intensity is appropriate and whether there are potential risks.

[0088] In other embodiments, the AI-based respiratory exercise system for patients with chronic obstructive pulmonary disease also includes a healthcare management platform. The healthcare provider communicates with the patient and AI analysis modules via a web application or dedicated application to achieve the following functions:

[0089] Intelligent Early Warning of Acute Exacerbation Risk: The AI ​​analysis module comprehensively assesses the level of acute exacerbation risk based on multi-dimensional data anomalies during patient training. The specific judgment logic is as follows:

[0090] Data monitoring dimensions: Within 3 consecutive days, the patient's peak expiratory flow rate decreased by ≥20% compared to their personal historical baseline; or the respiratory rate increased by ≥5 breaths / minute compared to the baseline; or the blood oxygen saturation decreased by ≥3% compared to the baseline and lasted for more than 10 minutes; or the dyspnea score (mMRC) in the patient's check-in record increased by 1 point or more compared to the baseline.

[0091] Risk level assessment:

[0092] Low risk (yellow alert): Only one of the above monitoring dimensions shows a mild abnormality, and there are no other accompanying symptoms. The AI ​​analysis module automatically generates home management guidelines, which are pushed through the patient's application. The guidelines include: increasing the number of daily training sessions to 2-3 times, increasing rest frequency, increasing fluid intake, using emergency inhalers as needed, and closely monitoring changes in symptoms. The system also automatically tracks the patient's status after 24 hours, and escalates the alert if the abnormality persists.

[0093] Medium Risk (Orange Alert): Two or more abnormalities are detected in the above monitoring dimensions, or a single abnormality is detected but the magnitude is significant (e.g., peak expiratory flow rate decreases by ≥30%). The AI ​​analysis module immediately sends an alert message to the healthcare provider, including basic patient information, details of the abnormal indicators, a training trend chart for the past 7 days, and the risk level assessment results. Upon receiving the alert, the healthcare provider reviews the patient data online, sends home management suggestions through the platform (e.g., adjusting medication regimens, increasing oxygen therapy time, and earlier follow-up appointments), and marks patients requiring special attention.

[0094] High Risk (Red Alert): Multiple serious abnormalities occur in the above monitoring dimensions, and the patient's check-in record shows typical acute exacerbation symptoms such as worsening shortness of breath, increased sputum volume, or yellow / green sputum color. The AI ​​analysis module immediately and automatically sends an emergency alert to the medical staff, and simultaneously issues a voice and text prompt through the patient's application: "Please contact a doctor immediately or go to the hospital for treatment." It also automatically generates an acute exacerbation assessment report containing all recent training data, physiological data, and symptom trends, and pushes it to the on-duty medical staff. After confirmation, the medical staff can proactively call the patient for remote guidance or arrange emergency medical treatment.

[0095] Remote medical management and intervention: The medical management platform establishes a personalized rehabilitation record for each patient, and medical staff can view the following content at any time:

[0096] Training records: flow rate curves, pass rate, replay of test ball 4 movement trajectory, training duration and frequency for each training session;

[0097] Rehabilitation indicators: weekly / monthly average peak expiratory flow trend, respiratory rate variability, blood oxygen saturation distribution map, and heart rate recovery time;

[0098] Changes in condition: CAT score change curve, records of acute exacerbations, medication adherence statistics, and patient complaint log entries.

[0099] Based on the above data, medical staff can perform the following operations online:

[0100] Remote adjustment of rehabilitation plan: Modify parameters such as patient flow threshold (F_low / F_high), single training duration, daily training frequency, and resistance adjustment gradient. The modified instructions are sent to the AI ​​control system and patient application in real time.

[0101] Push follow-up visit reminders: Automatically or manually generate follow-up visit reminders based on changes in the patient's condition, push them through the patient's application, and can be linked to the hospital's registration system;

[0102] Customized patient education content: Targeting patients' recent weaknesses in rehabilitation, we provide them with specialized patient education courses (such as "Video tutorials on the correct use of inhalers" and "Precautions for home oxygen therapy").

[0103] Online communication and follow-up: The platform's built-in instant messaging function allows for text and voice communication with patients, replacing some offline follow-ups and reducing the burden of patients traveling to and from the hospital.

[0104] Seamless Integration of Home-Based Rehabilitation and In-Hospital Management: The healthcare management platform connects with the hospital's HIS system. When patients visit the hospital, doctors can directly access complete data on the patient's home rehabilitation period, including training trend charts, risk warning records, and remote intervention history, providing objective evidence for treatment decisions. Simultaneously, information such as the patient's pulmonary function test results and treatment plan adjustments during hospitalization can also be synchronized to the patient's rehabilitation plan, ensuring continuity and consistency between in-hospital treatment and home-based rehabilitation.

[0105] The control module is also used to receive rehabilitation plan adjustment instructions from medical staff and update the adjusted parameters to the AI ​​control system to achieve remote rehabilitation management.

[0106] In other embodiments, the AI ​​analysis module also incorporates a risk warning sub-model. This sub-model uses an LSTM (Long Short-Term Memory) network to perform sequence modeling on the patient's multidimensional time-series data (flow rate, frequency, blood oxygen, heart rate, symptom check-in) from multiple consecutive training sessions, learning the patient's individualized physiological fluctuation patterns. When the real-time monitoring data sequence significantly deviates from the patient's historical baseline pattern, the risk warning sub-model automatically calculates a risk index and triggers a corresponding level of warning based on a preset threshold. As the number of training sessions increases, this sub-model continuously optimizes the fitting accuracy to the patient's individual characteristics through incremental learning, reducing false positives and false negatives.

[0107] This solution achieves dynamic adaptive adjustment of respiratory resistance. By monitoring components to collect data such as peak flow and waveform characteristics in real time, and combining them with an AI model to dynamically determine the patient's fatigue state, it effectively avoids the problems of excessive initial load and insufficient stimulation at the end caused by fixed resistance, thus improving the safety and comfort of the training process. At the same time, this solution uses the mechanical structure of the test airway 3, test ball 4, adjustment component and drive component in conjunction with AI control to automatically and smoothly adjust the effective air output cross-sectional area within a single respiratory cycle. This eliminates the need for patients to manually change gears or adjust valves, reducing the operational threshold and improving compliance with long-term pulmonary rehabilitation training.

[0108] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. An AI-based breathing exercise device for patients with chronic obstructive pulmonary disease, comprising a mouthpiece (1), characterized in that, It also includes an AI control system; one end of the air nozzle (1) is fixedly connected to a bend (2), and the other end of the bend (2) away from the air nozzle (1) is fixedly connected to a test airway (3), and a test ball (4) is slidably fitted inside the test airway (3); The test air passage (3) is provided with an adjustment component for adjusting the air output of the test air passage (3) at the end away from the bend (2), and a drive component for driving the adjustment component to operate. The test airway (3) is equipped with a monitoring component for monitoring the gas flow rate value inside the test airway (3); the driving component is used to drive the regulating component to adjust the gas output of the test airway (3) based on the flow rate value.

2. The AI-based respiratory training device for patients with chronic obstructive pulmonary disease according to claim 1, characterized in that, The adjustment component includes an air outlet (5), and a hexagonal groove (6) is opened on the side wall of the air outlet (5). A top plate (7) is rotatably fitted on the air outlet (5). Several guide grooves (8) are opened circumferentially on the top plate (7). Several triangular plates (9) are provided between the top plate (7) and the air outlet (5). A slider (10) is fixedly connected to each triangular plate (9). The slider (10) is slidably fitted with the slide groove (6). A slide rod (11) is also fixedly connected to each triangular plate (9). The slide rod (11) is slidably fitted with its adjacent guide groove (8). A ring gear (12) is sleeved on the top plate (7).

3. The AI-based respiratory training device for patients with chronic obstructive pulmonary disease according to claim 2, characterized in that, The drive assembly includes an electric cylinder (13) fixedly connected to the side wall of the air outlet (5). The AI ​​control system is used to control the extension and retraction of the output shaft of the electric cylinder (13). A rack (14) is fixedly connected to the output shaft of the electric cylinder (13), and the rack (14) meshes with the ring gear (12).

4. The AI-based respiratory exercise device for patients with chronic obstructive pulmonary disease according to claim 3, characterized in that, The monitoring components include a gas flow sensor fixedly connected to the inner wall of the test air passage (3), and an AI control system for receiving the gas flow value in the test air passage (3) sent by the gas flow sensor and controlling the extension and retraction of the output shaft of the electric cylinder (13) based on the gas flow value. The test airway (3) is equipped with an anti-choking component to prevent the test ball (4) from entering the patient's airway.

5. The AI-based respiratory exercise device for patients with chronic obstructive pulmonary disease according to claim 4, characterized in that, The anti-choking component includes a limiting block (15) that is symmetrically fixed to the inner wall of the test airway (3).

6. The AI-based respiratory training device for patients with chronic obstructive pulmonary disease according to claim 5, characterized in that, A handle (16) is symmetrically fixed to the side wall of the test airway (3).

7. An AI-based respiratory training system for patients with chronic obstructive pulmonary disease, applicable to the AI-based respiratory training device for patients with chronic obstructive pulmonary disease as described in any one of claims 1-6, characterized in that, include: The data acquisition module is used to collect exercise data during the patient's breathing exercises using a gas flow sensor and send the exercise data to the AI ​​analysis module. The AI ​​analysis module is used to input respiratory exercise sample data from patients with different COPD severity levels into a CNN neural network model for learning and training, thereby obtaining a respiratory function assessment model for COPD patients. The AI ​​analysis module is also used to input exercise data into the respiratory function assessment model for COPD patients, generate personalized exercise plans, and send the personalized exercise plans to the control module. The control module is used to convert personalized exercise programs into control commands and send control commands to the drive components and use the control commands to control the extension and retraction of the electric cylinder (13) to adjust the air output of the test airway (3).

8. The AI-based respiratory training system for patients with chronic obstructive pulmonary disease according to claim 7, characterized in that, In the data acquisition module, the exercise data includes gas flow rate and respiratory rate data.

9. The AI-based respiratory training system for patients with chronic obstructive pulmonary disease according to claim 8, characterized in that, In the AI ​​analysis module, the breathing exercise sample data includes respiratory flow, respiratory rate, respiratory duration, airway resistance adaptation parameters, and corresponding respiratory function assessment results.

10. The AI-based respiratory training system for patients with chronic obstructive pulmonary disease according to claim 9, characterized in that, The AI ​​analysis module includes a feedback optimization subunit, which sends the data from each breathing exercise as new sample data to the COPD patient respiratory function assessment model for incremental learning.

Images