A knee postoperative adaptive progressive rehabilitation training method based on pain detection

Abstract

The application relates to the technical field of knee joint rehabilitation, in particular to a knee joint postoperative adaptive gradual rehabilitation training method based on pain detection, which comprises the following steps: initializing a knee joint rehabilitation machine, and making a patient wear an electromyographic acquisition device and the knee joint rehabilitation machine; the patient drives the rehabilitation machine to actively bend the knee joint, and records knee joint active activity angle data; the rehabilitation machine performs passive knee joint flexion movement, and collects training data required by a pain detection model; preprocessing and feature extraction are performed, a classifier is trained, and the pain detection model is obtained; a rehabilitation training mode is selected to train the patient; after the patient completes all the training, the patient actively bends the knee joint again, and records the knee joint active activity angle data. The application utilizes surface electromyographic signals of the patient to perform pain detection, realizes automatic identification of a knee joint activity range, and performs multi-mode training according to the training condition of the patient, thereby relieving the problem of tight resources of rehabilitation doctors.

Description

Technical Field

[0001] This application relates to the field of knee joint rehabilitation technology for patients with periarticular fractures of the knee joint, and in particular to an adaptive progressive rehabilitation training method for knee joint surgery based on pain detection. Background Technology

[0002] Traffic accidents, workplace injuries, and the rising incidence of osteoporosis due to an aging population result in a large number of fracture patients in my country every year. Fracture treatment involves three basic stages: reduction, fixation, and rehabilitation. Reduction and fixation are usually performed surgically in hospitals, offering good completion rates and controllability. However, compared to reduction and fixation, my country faces a relative shortage of clinical rehabilitation therapists. Many patients cannot receive timely and effective rehabilitation training, or misconceptions such as the need for rest after a fracture lead to joint adhesions and swelling, severely limiting joint mobility and significantly hindering postoperative rehabilitation. In recent years, the development of rehabilitation robots has provided a new approach to address the problem of fracture patients missing the optimal time for rehabilitation training due to a lack of rehabilitation therapists. The advantages of robots in motion stability and repeatability allow them to partially replace rehabilitation therapists, assisting patients in high-intensity, repetitive exercise training, thereby helping them achieve better rehabilitation results.

[0003] In clinical rehabilitation training, especially for patients with joint adhesions and swelling who urgently require rehabilitation, the maximum range of motion for knee joint training should ideally be designed to induce mild pain. At this point, the adhered tissues can be appropriately stretched, thereby gradually increasing the range of motion of the joint. Currently, clinical confirmation relies on the patient's subjective expression and the rehabilitation physician's tactile experience, making objective quantification impossible.

[0004] However, current research on rehabilitation robots specifically designed for post-fracture rehabilitation training is limited. Existing rehabilitation robots are simple in function and cumbersome to operate. In particular, existing machines can only pre-set the range of motion of the knee joint and cannot adjust the range of motion according to the patient's individual condition because they cannot detect and monitor pain. This requires real-time operation by rehabilitation physicians, which does not alleviate the problem of limited rehabilitation resources. Therefore, to meet the post-operative rehabilitation needs of patients with peri-knee fractures, especially the early motor function rehabilitation training needs, there is an urgent need for a rehabilitation training method that can autonomously sense the patient's current optimal range of motion of the knee joint and adaptively control multi-modal training. Summary of the Invention

[0005] This application provides an adaptive progressive rehabilitation training method for knee joint surgery based on pain detection. By using the patient's surface electromyography signals for pain detection, it achieves automatic identification of the range of motion of the knee joint and conducts multi-mode training according to the patient's training status, thus alleviating the problem of the shortage of rehabilitation physician resources.

[0006] To address the aforementioned technical problems, this application provides a method for adaptive progressive rehabilitation training of the knee joint after surgery based on pain detection, comprising the following steps: First, the knee joint rehabilitation machine is initialized, and the patient wears the electromyography (EMG) acquisition device and the knee joint rehabilitation machine; then, after the patient wears the EMG acquisition device and the knee joint rehabilitation machine, the patient actively flexes the knee joint, recording the active knee joint angle data before training; next, the patient passively flexes the knee joint following the knee joint rehabilitation machine, collecting the training data required for the pain detection model; then, the collected training data is preprocessed and features are extracted, a support vector machine is selected as the classifier, and the extracted two-class sample features are used to train the classifier to obtain the pain detection model; next, a rehabilitation training mode is selected for the patient to train; the rehabilitation training modes include: passive movement training, pain-adaptive passive training, adversarial training, and pain-adaptive adversarial training; finally, after the patient completes all training, the patient again actively flexes the knee joint, recording the active knee joint angle data after training.

[0007] In some exemplary embodiments, the patient puts on the electromyography (EMG) acquisition device and the knee joint rehabilitation machine, including: Step 1, the patient puts on the EMG acquisition device, with the electrode pads of the EMG acquisition device attached to the rectus femoris, vastus medialis, long head of the biceps femoris, and semitendinosus muscles of the thigh; the EMG acquisition device is activated to begin acquiring surface EMG signal data; Step 2, the patient puts on the knee joint rehabilitation machine, ensuring that the bionic joint link coincides with the rotation center of the knee joint, and the gap between the patient's affected limb and the knee joint rehabilitation machine is filled with an airbag; Step 3, the patient fills in personal information on the system interface of the knee joint rehabilitation machine to establish a personal training record file.

[0008] In some exemplary embodiments, recording the active knee joint angle data before training includes: adjusting the knee joint rehabilitation machine to an unloaded state so that the knee joint rehabilitation machine can follow the flexion movement of the affected limb; the patient actively flexes the knee joint and relaxes after reaching the maximum knee joint flexion angle; the knee joint rehabilitation machine records the maximum knee joint angle during the patient's active knee joint flexion process.

[0009] In some exemplary embodiments, the collection of training data required for the pain detection model includes: Step 1, the knee rehabilitation machine moves the affected limb to an initial position, and gradually increases the knee angle from the initial position until the patient feels pain at a level of 3 on the digital pain rating scale, and records the knee angle at this time; Step 2, a trajectory curve is generated based on the recorded knee angle to perform knee flexion movement, and the maximum angle is maintained for 5 seconds; Step 3, the knee flexion movement in Step 2 is repeated 3 times, and the patient's electromyographic signal during the knee flexion movement is recorded as the training signal for the pain detection model, and pain and non-pain labels are added to the electromyographic signal.

[0010] In some exemplary embodiments, a pain detection model is obtained by training a classifier using extracted features from two classes of samples, including: first, preprocessing the signal; then, extracting features from the preprocessed electromyographic signal using a sliding window; finally, using a support vector machine as a classifier and training the classifier using the extracted features from the two classes of samples; optimizing the classifier hyperparameters using cross-validation and grid optimization; selecting the hyperparameter model with the highest evaluation index to complete the classifier training and establish the pain detection model; the classifier hyperparameters include cost and weight.

[0011] In some exemplary embodiments, the passive exercise training process includes: setting parameters before the passive exercise training begins; generating a corresponding motion trajectory after parameter confirmation; and starting the passive exercise training; the knee joint rehabilitation machine moving the affected limb passively according to the planned trajectory; holding the knee joint rehabilitation machine at the set termination angle for 10 seconds; returning the knee joint rehabilitation machine to the initial angle after 10 seconds; performing multiple training sessions according to the set number of exercise parameters; and saving the data from this training process in a folder after the training is completed.

[0012] In some exemplary embodiments, the process of pain-adaptive passive training includes: setting parameters before the start of pain-adaptive passive training; generating corresponding motion trajectories after parameter confirmation; and starting pain-adaptive passive motion training; the knee joint rehabilitation machine passively moves the affected limb according to the planned trajectory; if one of the following conditions is met, the knee joint rehabilitation machine enters the holding phase, maintaining the current angle for 10 seconds: First condition: the patient actively stops the motor movement due to excessive pain; Second condition: the motor movement angle reaches the termination angle; Third condition: the pain detection model outputs a judgment result of 1; After 10 seconds, the knee joint rehabilitation machine moves the affected limb back to the initial angle; a single rehabilitation training session. Upon completion, the motion frequency parameters will be updated based on the force-velocity ratio and resistance change rate during training. Multiple training sessions will be conducted based on the set number of repetitions. After each three flexion exercises, the range of motion will be updated based on the number of repetitions required to enter the hold phase. If the number of repetitions for the first hold phase is greater than or equal to 2, the termination angle will decrease by 3°. If the number of repetitions for the second hold phase is greater than or equal to 2, the termination angle will increase by 3°. If the number of repetitions for the third hold phase is greater than or equal to 2, the termination angle will remain unchanged. Training will continue based on the updated motion parameters. After training, the data from this training session will be saved in a folder.

[0013] In some exemplary embodiments, the process of resistance training includes: setting parameters before starting resistance training; generating a corresponding motion trajectory after parameter confirmation; passively moving the affected limb according to the planned trajectory; maintaining the knee joint rehabilitation machine at the set termination angle for 10 seconds and entering impedance control mode, where the patient performs isometric contraction training by flexing and extending the knee joint; returning the affected limb to the initial angle after 10 seconds; performing multiple training sessions according to the set number of movements; and saving the data from the training process in a folder after the training is completed.

[0014] In some exemplary embodiments, the process of pain adaptive resistance training includes: setting parameters before the start of pain adaptive resistance training; generating corresponding motion trajectories after parameter confirmation; and initiating pain adaptive resistance exercise training. The knee joint rehabilitation machine passively moves the affected limb according to the planned trajectory. If one of the following conditions is met, the device enters a holding phase, maintaining the current angle for 10 seconds and entering impedance control mode, where the patient performs isometric contraction training by flexing and extending the knee joint. The first condition is that the patient actively stops the motor movement due to excessive pain. The second condition is that the motor movement angle reaches the termination angle. The third condition is that the pain detection model outputs a judgment result of 1. After 10 seconds, the knee joint rehabilitation machine moves the affected limb back to the initial angle. After a single rehabilitation training session, the system will adjust the parameters according to the training process. The force-velocity ratio and resistance change rate are used to update the motion frequency parameters; the impedance control parameters are updated based on the maximum displacement difference, maximum velocity difference, and maximum electromyographic characteristic difference during the patient's impedance training; multiple training sessions are conducted based on the set number of motion repetitions; after each set of three flexion movements, the range of motion is updated based on the number of times each condition for entering the hold phase is met; if the number of times the first condition for entering the hold phase is greater than or equal to 2, the termination angle is decreased by 3°; if the number of times the second condition for entering the hold phase is greater than or equal to 2, the termination angle is increased by 3°; if the number of times the third condition for entering the hold phase is greater than or equal to 2, the termination angle remains unchanged; training continues based on the updated motion parameters; after training, the data from this training process is saved in a folder.

[0015] In some exemplary embodiments, recording the active knee joint angle data after patient training includes: adjusting the knee joint rehabilitation machine to an unloaded state so that the knee joint rehabilitation machine can follow the flexion movement of the affected limb; the patient actively flexes the knee joint and relaxes after reaching the maximum knee joint flexion angle; the knee joint rehabilitation machine records the maximum knee joint angle during the patient's active knee joint flexion process.

[0016] The technical solution provided in this application has at least the following advantages:

[0017] This application provides a method for adaptive progressive rehabilitation training of the knee joint after surgery based on pain detection, including the following steps: First, the knee joint rehabilitation machine is initialized, and the patient wears the electromyography (EMG) acquisition device and the knee joint rehabilitation machine; then, after the patient wears the EMG acquisition device and the knee joint rehabilitation machine, the patient drives the knee joint rehabilitation machine to perform active knee flexion, and the active knee joint angle data before training is recorded; next, the patient follows the knee joint rehabilitation machine to perform passive knee flexion movements, and the training data required for the pain detection model is collected; then, the collected training data is preprocessed and features are extracted, a support vector machine is selected as the classifier, and the extracted two-class sample features are used to train the classifier to obtain the pain detection model; next, a rehabilitation training mode is selected to train the patient; the rehabilitation training mode includes: passive movement training, pain-adaptive passive training, adversarial training, and pain-adaptive adversarial training; finally, after the patient completes all training, the patient drives the knee joint rehabilitation machine again to perform active knee flexion, and the active knee joint angle data after training is recorded.

[0018] The knee joint postoperative adaptive progressive rehabilitation training method based on pain detection provided in this application, on the one hand, achieves autonomous pain detection by analyzing and processing the patient's electromyographic signals based on the physiological phenomena exhibited by the human body when experiencing pain, thus classifying knee joint pain or non-pain states without relying on the patient's active reports or the doctor's experience. Currently, clinical rehabilitation can only determine whether an appropriate level of pain has been reached based on the patient's feedback and the doctor's experience, and this is used as the patient's knee joint's limit of range of motion, which is highly subjective. The pain state detection technology mentioned in this application uses the patient's electromyographic signals as the analysis object, which can more objectively determine the patient's pain state and conduct rehabilitation exercises with an appropriate range of knee joint motion, thereby improving the effectiveness of rehabilitation training.

[0019] On the other hand, this application adaptively adjusts the training termination angle based on the conditions for each entry into the holding phase during rehabilitation training, in order to adapt to changes in the patient. Because the affected area of ​​the knee joint is loosened, the pain threshold increases, and the range of motion of the knee joint expands during rehabilitation training, the rehabilitation machine needs to be adjusted according to the patient's condition each time to facilitate more effective rehabilitation training.

[0020] Furthermore, this application designs passive motion training, pain-adaptive passive training, resistance training, and pain-adaptive resistance training based on common clinical knee joint rehabilitation techniques and combined with pain detection methods. This training method can meet the requirements of post-operative rehabilitation training for knee fractures, helping patients stretch the joint to relieve joint adhesions and also helping them strengthen muscles and restore muscle strength. Attached Figure Description

[0021] One or more embodiments are illustrated by way of example with reference to the accompanying drawings. These illustrations do not constitute a limitation on the embodiments, and unless otherwise stated, the figures in the drawings are not to be limited by scale.

[0022] Figure 1 This is a flowchart illustrating an adaptive progressive rehabilitation training method for knee joint surgery based on pain detection, provided in an embodiment of this application.

[0023] Figure 2 A flowchart illustrating the training data required for the pain detection model to be acquired by the knee joint rehabilitation machine provided in this application embodiment.

[0024] Figure 3 This is a flowchart illustrating the training of the pain detection model provided in an embodiment of this application.

[0025] Figure 4 A flowchart of rehabilitation training provided for an embodiment of this application. Detailed Implementation

[0026] As can be seen from the background technology, there is currently little research on rehabilitation robots that meet the needs of post-fracture rehabilitation training. Existing post-fracture rehabilitation machines have simple functions and are cumbersome to operate.

[0027] To address the aforementioned technical problems, this application provides a method for adaptive progressive rehabilitation training of the knee joint after surgery based on pain detection, comprising the following steps: First, the knee joint rehabilitation machine is initialized, and the patient wears the electromyography (EMG) acquisition device and the knee joint rehabilitation machine; then, after the patient wears the EMG acquisition device and the knee joint rehabilitation machine, the patient actively flexes the knee joint, recording the active knee joint angle data before training; next, the patient passively flexes the knee joint following the knee joint rehabilitation machine, collecting the training data required for the pain detection model; then, the collected training data is preprocessed and features are extracted, a support vector machine is selected as the classifier, and the extracted two-class sample features are used to train the classifier to obtain the pain detection model; next, a rehabilitation training mode is selected for the patient to train; the rehabilitation training modes include: passive movement training, pain-adaptive passive training, adversarial training, and pain-adaptive adversarial training; finally, after the patient completes all training, the patient again actively flexes the knee joint, recording the active knee joint angle data after training. This application provides an adaptive progressive rehabilitation training method for knee joint surgery based on pain detection. By using the patient's surface electromyography signals for pain detection, it achieves automatic identification of the range of motion of the knee joint and conducts multi-mode training according to the patient's training status, thus alleviating the problem of the shortage of rehabilitation physician resources.

[0028] The embodiments of this application will now be described in detail with reference to the accompanying drawings. However, those skilled in the art will understand that many technical details have been provided in the embodiments of this application to facilitate a better understanding of the application. However, the technical solutions claimed in this application can be implemented even without these technical details and various variations and modifications based on the following embodiments.

[0029] See Figure 1 This application provides a method for adaptive progressive rehabilitation training of the knee joint after surgery based on pain detection, including the following steps:

[0030] Step S1: Initialize the knee rehabilitation machine and have the patient put on the electromyography acquisition device and the knee rehabilitation machine.

[0031] Step S2: After the patient puts on the electromyography (EMG) acquisition device and the knee joint rehabilitation machine, the patient actively flexes the knee joint and records the active range of motion of the knee joint before training.

[0032] Step S3: The patient follows the knee joint rehabilitation machine to perform passive knee flexion movements, and the training data required for the pain detection model is collected.

[0033] Step S4: Preprocess and extract features from the collected training data, select support vector machine as classifier, and train the classifier using the extracted features of the two classes of samples to obtain the pain detection model.

[0034] Step S5: Select a rehabilitation training mode to train the patient; rehabilitation training modes include: passive movement training, pain-adaptive passive training, resistance training, and pain-adaptive resistance training.

[0035] Step S6: After the patient completes all the training, the knee joint rehabilitation machine is used again to perform active knee flexion, and the active range of motion of the knee joint after training is recorded.

[0036] This application provides a pain detection-based adaptive progressive rehabilitation training method for knee joint postoperative patients. The method includes: initializing the rehabilitation machine; the patient wearing the rehabilitation device and filling in patient information; recording the active range of motion of the knee joint before training; collecting training data required for the pain detection model; preprocessing and extracting features from the training data; selecting a support vector machine as the classifier for training to obtain the pain detection model; the patient selecting a rehabilitation training mode for training; and recording the active range of motion of the knee joint after the patient completes all training. This application utilizes a pain detection method based on electromyography (EMG) signals to design a multi-mode training method for a rehabilitation machine capable of autonomously sensing the range of motion of the knee joint, enabling patients to perform rehabilitation exercises with an appropriate range of motion, thereby improving the effectiveness of rehabilitation training.

[0037] In some embodiments, step S1, which involves having the patient wear the electromyography (EMG) acquisition device and the knee joint rehabilitation machine, includes the following steps:

[0038] Step 1: The patient wears the electromyography (EMG) acquisition device, with the electrodes of the device attached to the rectus femoris, vastus medialis, long head of the biceps femoris, and semitendinosus muscles of the thigh. The EMG acquisition device is then activated to begin acquiring surface EMG signal data.

[0039] Step 2: The patient puts on the knee joint rehabilitation machine, ensuring that the bionic joint link coincides with the rotation center of the knee joint, and the gap between the patient's affected limb and the knee joint rehabilitation machine is filled by an airbag.

[0040] Step 3: Fill in your personal information on the system interface of the knee joint rehabilitation machine to create a personal training record file for the patient.

[0041] In some embodiments, step S2, recording the active knee joint angle data before training, includes: adjusting the knee joint rehabilitation machine to an unloaded state so that the knee joint rehabilitation machine can follow the flexion movement of the affected limb; the patient actively flexes the knee joint and relaxes after reaching the maximum knee joint flexion angle; the knee joint rehabilitation machine records the maximum knee joint angle during the patient's active knee joint flexion process.

[0042] Figure 2 This document illustrates a flowchart of the process for collecting training data required for a pain detection model using a knee rehabilitation machine provided in an embodiment of this application. Figure 2 As shown, in some embodiments, step S3 involves collecting the training data required for the pain detection model, including the following steps:

[0043] Step 1: The knee joint rehabilitation machine moves the affected limb to the initial position. From the initial position, the knee joint angle is gradually increased until the patient feels pain at a level of 3 on the pain rating scale. The knee joint angle at this time is recorded.

[0044] Step 2: Generate a trajectory curve based on the recorded knee joint angle, perform knee flexion exercises, and hold the maximum angle for 5 seconds.

[0045] Step 3: Repeat the knee flexion exercise from Step 2 three times. Record the patient's electromyographic (EMG) signals during the knee flexion exercise as training signals for the pain detection model, and add pain and non-pain labels to the EMG signals.

[0046] Figure 3 The flowchart for training a pain detection model is shown. Figure 3 As shown, in some embodiments, the classifier is trained using the extracted features from the two classes of samples to obtain a pain detection model, including the following steps:

[0047] Step 1: Preprocess the signal.

[0048] Step 2: Extract the features of the preprocessed electromyographic signals using a sliding window method.

[0049] Step 3: Use a support vector machine as a classifier and train the classifier using the extracted features from the two classes of samples; optimize the hyperparameters of the classifier using cross-validation and grid optimization; select the hyperparameter model with the highest evaluation index to complete the classifier training and establish a pain detection model; the classifier hyperparameters include cost and weight.

[0050] Specifically, the signal preprocessing steps in step one include: performing a 10Hz-500Hz bandpass filter on the signal to remove 50Hz power frequency interference; and removing useless data from the original data based on the markers in the signal, while retaining the electromyographic signals from the beginning to the end of each rehabilitation training session.

[0051] Specifically, step two mainly involves feature extraction. Features of the preprocessed electromyographic signal are extracted using a sliding window method. The extracted features include mean absolute value (MAV), root mean square (RMS), peak-to-peak value (PP), standard deviation (STD), wavelength (WL), number of zero crossings (ZC), mean frequency (MNF), median frequency (MDF), mean power (MNP), and approximate entropy (ApEn). The extracted features are then regularized.

[0052] Figure 4 A flowchart of rehabilitation training is shown. (For example...) Figure 4As shown, in some embodiments, step S5 involves selecting a rehabilitation training mode to train the patient. The passive exercise training process includes: setting parameters before starting passive exercise training; generating a corresponding motion trajectory after parameter confirmation; starting passive exercise training; the knee joint rehabilitation machine moving the affected limb passively according to the planned trajectory; holding the knee joint rehabilitation machine at the set termination angle for 10 seconds; returning the knee joint rehabilitation machine to the initial angle after 10 seconds; performing multiple training sessions according to the set number of exercise parameters; and saving the data from this training process in a folder after the training is completed.

[0053] In some embodiments, step S5 selects a rehabilitation training mode to train the patient. The process of pain adaptive passive training includes: setting parameters before the start of pain adaptive passive training, generating a corresponding motion trajectory after parameter confirmation, and starting pain adaptive passive motion training; the knee joint rehabilitation machine drives the affected limb to perform passive movement according to the planned trajectory; if one of the following conditions is met, the knee joint rehabilitation machine enters the holding phase and holds the current angle for 10 seconds.

[0054] In the first scenario, the patient actively stops the motor due to excessive pain.

[0055] In the second scenario, the motor's movement angle reaches the termination angle.

[0056] In the third scenario, the pain detection model outputs a judgment result of 1.

[0057] Ten seconds later, the knee joint rehabilitation machine returns the affected limb to its initial angle; after a single rehabilitation training session, the motion frequency parameter will be updated according to formula (1) based on the force-velocity ratio and resistance change rate during the training process:

[0058]

[0059] Where r is the actual frequency of rehabilitation; k r1 ,k r2 The compensation coefficient is r0; the initial setting frequency for rehabilitation is r0; F(t) is the human-computer interaction force; and v(t) is the motor speed. It is the rate of change of human-computer interaction force.

[0060] The training is performed multiple times based on the set number of repetitions; after each three flexion exercises are completed, the range of motion is updated based on the number of times each condition for entering the hold phase is met.

[0061] If the condition for entering the holding phase is that the number of times the first case is greater than or equal to 2, then the termination angle is reduced by 3°.

[0062] If the condition for entering the holding phase is that the number of times the second case is greater than or equal to 2, then the termination angle increases by 3°.

[0063] If the condition for entering the holding phase is that the number of times the third case is greater than or equal to 2, then the termination angle remains unchanged.

[0064] Continue training based on the updated motion parameters; after training, save the data from this training process in a folder.

[0065] In some embodiments, step S5 involves selecting a rehabilitation training mode to train the patient. The process of resistance exercise training includes: setting parameters before starting the resistance exercise training; generating a corresponding motion trajectory after parameter confirmation; and starting the resistance exercise training; the knee joint rehabilitation machine passively moving the affected limb according to the planned trajectory; maintaining the knee joint rehabilitation machine at the set termination angle for 10 seconds and entering the impedance control mode, where the patient performs isometric contraction training by flexing and extending the knee joint; returning the affected limb to the initial angle after 10 seconds; performing multiple training sessions according to the set number of movements; and saving the data from this training process in a folder after the training is completed.

[0066] In some embodiments, step S5 selects a rehabilitation training mode to train the patient. The process of pain adaptive resistance training includes: setting parameters before the start of pain adaptive resistance training; generating corresponding motion trajectories after parameter confirmation; and starting pain adaptive resistance exercise training. The knee joint rehabilitation machine drives the affected limb to perform passive movements according to the planned trajectory. If one of the following conditions is met, the device enters the holding phase, holds the current angle for 10 seconds, and enters the impedance control mode, where the patient flexes and extends the knee joint to perform isometric contraction training.

[0067] The first scenario: The patient actively stops the motor operation due to excessive pain.

[0068] The second scenario: The motor's movement angle reaches the termination angle.

[0069] The third scenario: The pain detection model outputs a judgment result of 1.

[0070] Ten seconds later, the knee joint rehabilitation machine will return the affected limb to its initial angle; after a single rehabilitation training session, the movement frequency parameters will be updated based on the force-speed ratio and resistance change rate during the training process.

[0071] The impedance control parameters are updated according to formula (2) based on the maximum displacement difference, maximum velocity difference and maximum electromyographic characteristic difference during the patient's impedance training process.

[0072]

[0073] Where, k s ,k v ,k f ,k eK is the compensation coefficient; d B d ,T ff For impedance control parameters; K d0 B d0 ,T ff0 Set the initial impedance control parameters: s0, v0, F0, RMS EMG0 These represent the maximum displacement, maximum velocity, maximum interaction force, and maximum RMS amplitude of electromyography during the initial resistance training process; s n ,v n ,F n RMS EMGn The maximum displacement, maximum velocity, maximum interaction force, and maximum electromyographic RMS amplitude during the nth resistance training session are given.

[0074] A fatigue level estimation method is established using formula (3). Where τ ave τ0 represents the average time taken for three consecutive training exercises, while τ0 represents the average time taken for the initial three training exercises. The fatigue level will be updated after every three training exercises, and the resistance training coefficient will be updated after each training session.

[0075]

[0076] The training is performed multiple times based on the set number of repetitions; after each three flexion exercises are completed, the range of motion is updated based on the number of times each condition for entering the hold phase is met.

[0077] If the condition for entering the holding phase is that the number of times the first case is greater than or equal to 2, then the termination angle is reduced by 3°.

[0078] If the condition for entering the holding phase is that the number of times the second case is greater than or equal to 2, then the termination angle increases by 3°.

[0079] If the condition for entering the holding phase is that the number of times the third case is greater than or equal to 2, then the termination angle remains unchanged.

[0080] Continue training based on the updated motion parameters; after training, save the data from this training process in a folder.

[0081] In some embodiments, step S6, recording the active knee joint angle data after training, includes: adjusting the knee joint rehabilitation machine to an unloaded state so that the knee joint rehabilitation machine can follow the flexion movement of the affected limb; the patient actively flexes the knee joint until the maximum knee joint flexion angle is reached and then relaxes; the knee joint rehabilitation machine records the maximum knee joint angle during the patient's active knee joint flexion process.

[0082] Based on the above technical solutions, this application provides a method for adaptive progressive rehabilitation training of the knee joint after surgery based on pain detection, including the following steps: First, the knee joint rehabilitation machine is initialized, and the patient wears the electromyography (EMG) acquisition device and the knee joint rehabilitation machine; then, after the patient wears the EMG acquisition device and the knee joint rehabilitation machine, the patient drives the knee joint rehabilitation machine to perform active knee flexion, and records the active knee joint angle data before training; next, the patient follows the knee joint rehabilitation machine to perform passive knee flexion movements, and collects the training data required for the pain detection model; then, the collected training data is preprocessed and features are extracted, a support vector machine is selected as the classifier, and the extracted two-class sample features are used to train the classifier to obtain the pain detection model; next, a rehabilitation training mode is selected to train the patient; the rehabilitation training mode includes: passive movement training, pain-adaptive passive training, adversarial training, and pain-adaptive adversarial training; finally, after the patient completes all training, the patient drives the knee joint rehabilitation machine again to perform active knee flexion, and records the active knee joint angle data after training.

[0083] The knee joint postoperative adaptive progressive rehabilitation training method based on pain detection provided in this application, on the one hand, achieves autonomous pain detection by analyzing and processing the patient's electromyographic signals based on the physiological phenomena exhibited by the human body when experiencing pain, thus classifying knee joint pain or non-pain states without relying on the patient's active reports or the doctor's experience. Currently, clinical rehabilitation can only determine whether an appropriate level of pain has been reached based on the patient's feedback and the doctor's experience, and this is used as the patient's knee joint's limit of range of motion, which is highly subjective. The pain state detection technology mentioned in this application uses the patient's electromyographic signals as the analysis object, which can more objectively determine the patient's pain state and conduct rehabilitation exercises with an appropriate range of knee joint motion, thereby improving the effectiveness of rehabilitation training.

[0084] On the other hand, this application adaptively adjusts the training termination angle based on the conditions for each entry into the holding phase during rehabilitation training to adapt to changes in the patient. Because the affected area of ​​the knee joint is loosened during rehabilitation training, the pain threshold increases, and the range of motion of the knee joint expands, the rehabilitation machine needs to be adjusted according to the patient's condition each time to facilitate more effective rehabilitation training. Furthermore, this application designs passive motion training, pain-adaptive passive training, resistance training, and pain-adaptive resistance training based on common clinical knee joint rehabilitation training techniques and combined with pain detection methods. This training method can meet the requirements of postoperative rehabilitation training for knee fractures, helping patients stretch the joint to relieve joint adhesions and also helping them strengthen muscles and restore muscle strength.

[0085] Those skilled in the art will understand that the above-described embodiments are specific examples of implementing this application, and in practical applications, various changes in form and detail may be made without departing from the spirit and scope of this application. Any person skilled in the art can make their own modifications and alterations without departing from the spirit and scope of this application; therefore, the scope of protection of this application should be determined by the scope defined in the claims.

Claims

1. A knee postoperative adaptive progressive rehabilitation training system based on pain detection, characterized in that, The system comprises a knee joint rehabilitation machine, an electromyography collection device, a data processing unit and a control unit, and is configured to perform the following control method: The knee joint rehabilitation machine is initialized, and the patient wears the electromyography collection device and the knee joint rehabilitation machine; After the patient wears the electromyography collection device and the knee joint rehabilitation machine, the knee joint rehabilitation machine is driven to perform active knee flexion, and the active movement angle data of the knee joint of the patient before training is recorded; The patient follows the knee joint rehabilitation machine to perform passive knee flexion movement, and collects the training data required by the pain detection model; The collected training data is preprocessed and feature extracted, a support vector machine is selected as a classifier, and the extracted two types of sample features are used to train the classifier to obtain the pain detection model; The patient is trained in the rehabilitation training mode; The rehabilitation training mode includes passive movement training, pain adaptive passive training, antagonistic training and pain adaptive antagonistic training; After the patient completes all the training, the knee joint rehabilitation machine is driven again to perform active knee flexion, and the active movement angle data of the knee joint of the patient after training is recorded; The process of the pain adaptive passive training includes: Before the pain adaptive passive training starts, parameters are set first, and after the parameters are confirmed, the corresponding movement trajectory is generated, and the pain adaptive passive movement training starts; the knee joint rehabilitation machine drives the affected limb to perform passive movement according to the planned trajectory; if one of the following conditions is met, the knee joint rehabilitation machine enters the holding stage, and the current angle is maintained for 10 seconds; The first condition is that the patient actively stops the motor movement due to strong pain sensation; The second condition is that the motor movement angle reaches the termination angle; The third condition is that the pain detection model outputs a judgment result of 1, indicating that the model detects pain; After 10 seconds, the knee joint rehabilitation machine drives the affected limb back to the initial angle; after a single rehabilitation training is completed, the movement frequency parameter is updated according to the force-speed ratio and the resistance change rate in the training process; multiple training is performed according to the set movement number parameter; every time three flexion movements are completed, the movement range is updated according to the number of times of entering the holding stage under each condition; When the number of times of entering the holding stage under the first condition is greater than or equal to 2, the termination angle is reduced by 3°; When the number of times of entering the holding stage under the second condition is greater than or equal to 2, the termination angle is increased by 3°; When the number of times of entering the holding stage under the third condition is greater than or equal to 2, the termination angle remains unchanged; According to the updated movement parameters, the training is continued; after the training is completed, the data in the training process is saved in a folder.

2. The pain detection based post-operative rehabilitation training system for knee joint as claimed in claim 1, wherein, The patient wears the electromyography collection device and the knee joint rehabilitation machine, including: Step 1: The patient wears the electromyography collection device, and the electrode pieces of the electromyography collection device are attached to the rectus femoris muscle, the vastus medialis muscle, the long head of the biceps femoris muscle and the semitendinosus muscle; the electromyography collection device is started, and surface electromyography signal data collection begins; Step 2: The patient wears the knee joint rehabilitation machine, and ensures that the bionic joint connecting rod coincides with the rotation center of the knee joint during wearing, and fills the gap between the affected limb of the patient and the knee joint rehabilitation machine through the air bag; Step 3: Fill in your personal information on the system interface of the knee joint rehabilitation machine to create a personal training record file for the patient.

3. The pain detection based post-operative rehabilitation training system for knee joint as claimed in claim 1 wherein, Record the patient's active knee joint range of motion data before training, including: Adjust the knee joint rehabilitation machine to the unloaded state so that the knee joint rehabilitation machine can follow the flexion movement of the affected limb; The patient actively flexes their knee joint until they reach the maximum flexion angle, then relaxes. The knee rehabilitation machine records the maximum angle of the knee joint during the patient's active knee flexion process.

4. The pain detection based post-operative knee rehabilitation training system according to claim 1, wherein, The training data required to collect the pain detection model includes: Step 1: The knee joint rehabilitation machine moves the affected limb to the initial position. From the initial position, the knee joint angle is gradually increased until the patient feels pain at a level of 3 on the pain rating scale. The knee joint angle at this time is recorded. Step 2: Generate a trajectory curve based on the recorded knee joint angle, perform knee flexion exercises, and hold the maximum angle for 5 seconds; Step 3: Repeat the knee flexion exercise from Step 2 three times. Record the patient's electromyographic (EMG) signals during the knee flexion exercise as training signals for the pain detection model, and add pain and non-pain labels to the EMG signals.

5. The pain detection based post-operative knee rehabilitation training system according to claim 1, wherein, The classifier is trained using the extracted features from the two classes of samples to obtain a pain detection model, including: Preprocess the signal; Features of preprocessed electromyographic signals were extracted using a sliding window method; Support vector machines are used as classifiers, and the classifiers are trained using the extracted features of two classes of samples. The hyperparameters of the classifiers are optimized using cross-validation and grid optimization. The hyperparameter model with the highest evaluation index is selected to complete the classifier training and establish a pain detection model. The hyperparameters of the classifiers include cost and weight.

6. The pain detection based post-operative knee rehabilitation training system according to claim 1, wherein, The process of passive exercise training includes: Before starting passive exercise training, parameters are set first. After the parameters are confirmed, the corresponding motion trajectory is generated, and passive exercise training begins. The knee joint rehabilitation machine moves the affected limb passively according to the planned trajectory. When the knee joint rehabilitation machine reaches the set termination angle, it is held for 10 seconds. After 10 seconds, the knee joint rehabilitation machine returns to the initial angle. Multiple training sessions are performed according to the set number of movements. After the training is completed, the data from this training process is saved in a folder.

7. The pain detection based post-operative knee rehabilitation training system as claimed in claim 1, wherein, The process of competitive sports training includes: Before starting the resistance training, parameters are first set. After parameter confirmation, a corresponding motion trajectory is generated, and the resistance training begins. The knee joint rehabilitation machine passively moves the affected limb according to the planned trajectory. After the knee joint rehabilitation machine reaches the set termination angle, it holds for 10 seconds and enters the resistance control mode, where the patient performs isometric contraction training by flexing and extending the knee joint. After 10 seconds, the knee joint rehabilitation machine moves the affected limb back to the initial angle. Multiple training sessions are performed according to the set number of repetitions. After the training is completed, the data from this training process is saved in a folder.

8. The pain detection based post-operative knee rehabilitation training system as claimed in claim 1, wherein, The process of pain adaptive adversarial training includes: Firstly, parameters are set before the pain adaptive antagonistic training starts. After the parameters are confirmed, corresponding movement trajectories are generated, and the pain adaptive antagonistic movement training starts. The knee rehabilitation machine drives the affected limb to perform passive movement according to the planned trajectory. If one of the following conditions is met, the device enters the holding stage, maintains the current angle for 10 seconds, and enters the impedance control mode. The patient performs isometric contraction training by flexing and extending the knee joint. The first condition: the patient actively stops the motor movement due to strong pain sensation. The second condition: the motor movement angle reaches the termination angle. The third condition: the pain detection model output judgment result is 1, indicating that the model detects pain. After 10 seconds, the knee rehabilitation machine drives the affected limb back to the initial angle. After a single rehabilitation training is completed, the movement frequency parameter is updated according to the force-speed ratio and resistance change rate in the training process. The impedance control parameter is updated according to the maximum displacement difference, maximum speed difference, and maximum muscle electrical characteristic difference in the patient's impedance training process. Multiple training is performed according to the set movement frequency parameter. Whenever three flexion movements are completed, the movement range is updated according to the number of times each condition enters the holding stage. When the number of times the first condition enters the holding stage is greater than or equal to 2, the termination angle is reduced by 3°. When the number of times the second condition enters the holding stage is greater than or equal to 2, the termination angle is increased by 3°. When the number of times the third condition enters the holding stage is greater than or equal to 2, the termination angle remains unchanged. According to the updated movement parameters, the training continues. After the training is completed, the data in this training process is saved in a folder.

9. The pain detection based post-operative knee rehabilitation training system as claimed in claim 1, wherein, Record the active activity angle data of the patient's knee joint after training, including: Adjust the knee rehabilitation machine to the unloading state, so that the knee rehabilitation machine can follow the flexion movement of the affected limb. The patient actively performs knee flexion, and relaxes after reaching the maximum angle of the patient's knee flexion. The knee rehabilitation machine records the maximum angle of the knee joint during the patient's active knee flexion.

Images