Orthopedic postoperative rehabilitation training method based on biomechanical feedback

Abstract

The invention discloses an orthopedic postoperative rehabilitation training method based on biomechanical feedback. The method comprises the following steps: synchronously collecting joint kinematics, myoelectricity and plantar pressure multi-source data of a patient; performing time sequence alignment and noise suppression on the data; constructing an individualized biomechanical function evaluation model, and generating a dynamic score covering the joint motion range, the muscle symmetry, the gait stability and the load tolerance; the training intensity, the auxiliary support and the rest time are adjusted in a self-adaptive mode based on the deviation between the score and the stage target; voice with clinical semantics and visual feedback are generated; and the evaluation and regulation strategy is continuously optimized through the closed-loop learning module. According to the technical scheme, precision, individualization and intelligentization of rehabilitation training are achieved, the recovery efficiency is effectively improved, and the risk of secondary injury is reduced.

Description

Technical Field

[0001] This invention relates to the medical field, and in particular to a postoperative rehabilitation training method for orthopedic surgery based on biomechanical feedback. Background Technology

[0002] With the continuous advancement of orthopedic surgical techniques, the importance of postoperative rehabilitation training in the functional recovery process of patients is becoming increasingly prominent. Modern rehabilitation medicine emphasizes individualized, dynamic, and precise training strategies to promote the coordinated repair of the skeletal, muscular, and joint systems. However, traditional rehabilitation programs rely heavily on standardized training manuals and physician experience, lacking an objective perception and feedback mechanism of the patient's real-time biomechanical state. This makes it difficult to achieve precise control of training intensity, angle, and frequency, thus affecting rehabilitation outcomes and potentially causing secondary injuries.

[0003] Among them, biomechanical feedback-based rehabilitation training methods aim to collect patients' kinematic and dynamic parameters, such as joint range of motion, muscle strength output, and gait symmetry, through sensors to construct individualized training response models. The core of this method lies in transforming physiological signals into quantifiable training guidance, enabling the rehabilitation process to adaptively adjust. However, existing feedback systems generally suffer from problems such as limited data dimensions, delayed response, and a lack of closed-loop control mechanisms.

[0004] Existing technologies typically focus only on the motion parameters of a single joint or local muscle group, neglecting the biomechanical coupling effects in multi-joint coordinated movements, leading to a disconnect between training movements and actual functional needs. Furthermore, most systems use static thresholds to determine training effectiveness, failing to dynamically adjust feedback strategies based on patient fatigue levels or tissue healing stages. In addition, the lack of semantic mapping between biomechanical data and clinical rehabilitation goals makes it difficult for systems to generate medically interpretable training recommendations. Therefore, in complex orthopedic postoperative scenarios, there is an urgent need for a rehabilitation training method that can integrate multi-source biomechanical information, achieve dynamic closed-loop regulation, and possess clinical interpretability. Summary of the Invention

[0005] The purpose of this invention is to provide a postoperative rehabilitation training method for orthopedic surgery based on biomechanical feedback, which solves the problems mentioned in the background art.

[0006] This invention is implemented as follows: a biomechanical feedback-based orthopedic postoperative rehabilitation training method, comprising the following specific steps:

[0007] Step 1: Collect multi-source biomechanical data of the patient during rehabilitation training. Simultaneously acquire the three-dimensional motion angle, angular velocity, acceleration of the target joint and the electromyography signal intensity of the relevant muscle groups through a wearable inertial measurement unit and surface electromyography sensor array. At the same time, use a plantar pressure distribution sensor to record the pressure center trajectory, stance phase duration and bilateral load distribution ratio in the gait cycle, forming a raw data stream covering kinematic, dynamic and physiological dimensions.

[0008] Step 2: Perform time alignment and noise suppression processing on the multi-source biomechanical data. Use wavelet transform combined with adaptive Kalman filtering algorithm to correct baseline drift and filter high-frequency interference in electromyography signals. Perform spline interpolation on kinematic data to compensate for time misalignment caused by signal transmission delay, and ensure that the synchronization accuracy of each modality data on the time axis is better than 5 milliseconds.

[0009] Step 3: Construct an individualized biomechanical function assessment model. Based on the preset rehabilitation stage division criteria, the processed biomechanical data is mapped to four core assessment dimensions: joint range of motion recovery index, muscle activation symmetry index, gait stability index, and load tolerance growth rate. A dynamic biomechanical function score is generated through a weighted fusion algorithm, where the weight of each dimension is personalized according to the type of surgery, postoperative days, and the patient's basic physical fitness parameters.

[0010] Step 4: Establish an adaptive adjustment mechanism for training parameters. Analyze the deviation between the biomechanical function score and the preset phased rehabilitation goals. When the score is below 90% of the target threshold, automatically reduce the number of repetitions of the training movements to 70% of the original plan and increase the auxiliary support intensity by 15% to 25%. When the score is consistently above 110% of the target threshold for three consecutive assessments, automatically increase the movement resistance torque by 10% to 20% and shorten the rest time between sets by 15 to 30 seconds.

[0011] Step 5: Generate feedback instructions with clinical semantic interpretation, and transform the control decision into voice and visual prompts containing medical expressions such as "the main reason for the current limited joint movement is the delayed activation of the quadriceps femoris" or "gait asymmetry is due to insufficient support of the arch of the affected side", and push them to the patient and rehabilitation physician in real time through the intelligent interactive terminal.

[0012] Step 6: Construct a closed-loop learning optimization module. Store the complete data sequence of each training session, the record of regulatory decisions, and the annotation of physician interventions in the local database. Use an incremental support vector machine model to perform pattern mining on the historical data and regularly update the parameter mapping rules in the biomechanical functional assessment model to improve the system's accuracy in predicting individual recovery trajectories and the adaptability of regulatory strategies.

[0013] Preferably, in step 1, the wearable inertial measurement unit operates at a sampling frequency of 200 frames per second. The sensors are arranged at the distal femur, proximal tibia, and pelvic surface landmarks. The real-time angular changes of the three degrees of freedom of knee joint flexion-extension, adduction-abduction, and internal-external rotation are calculated by rigid body kinematics inverse kinematics algorithm, with a measurement error of less than 1.2 degrees.

[0014] Preferably, the surface electromyography sensor array covers four key muscle groups: rectus femoris, vastus medialis, vastus lateralis, and semitendinosus. The electrode spacing is 20 mm, the signal gain is set to 1000 times, the common-mode rejection ratio is greater than 110 dB, and the skin-electrode impedance is monitored in real time during the acquisition process. When the impedance value exceeds 5 kΩ, a wearing status alarm is triggered.

[0015] Preferably, the plantar pressure distribution sensing device is integrated inside an adjustable-hardness rehabilitation insole, containing 128 piezoresistive sensing units, with a spatial resolution of 1.5 cm x 1.5 cm, a pressure measurement range of 0 to 1200 kPa, a sampling frequency of 100 Hz, and supporting dynamic recognition of 6 basic activity modes such as standing, walking, and climbing stairs.

[0016] Preferably, in step 3, the joint range of motion recovery index is calculated based on the decay rate of the difference between the maximum active flexion angle of the affected and healthy joints. It is set that the daily decay should be greater than 5% from the 14th to the 28th day after surgery. If the decay is less than 3% for 3 consecutive days, it is marked as delayed recovery.

[0017] Preferably, the muscle activation symmetry index is determined by calculating the ratio of the integral electromyographic signals of the same muscle groups on both sides during the same movement phase. The normal range is 0.85 to 1.15. When the ratio is lower than 0.7 or higher than 1.3, a muscle synergy training program is initiated.

[0018] Preferably, the gait stability index is based on a comprehensive score of the path length of the pressure center trajectory, the velocity variation coefficient, and the difference in load distribution between the left and right sides, with weights of 40%, 35%, and 25%, respectively. When the score is below 60, the balance training mode is automatically activated.

[0019] Preferably, the pre-set phased rehabilitation goals in step 4 are set according to internationally accepted orthopedic rehabilitation guidelines, and are divided into three phases: inflammation resolution phase, tissue remodeling phase, and functional enhancement phase. Each phase lasts for 14 days, and the corresponding biomechanical function score benchmarks for each phase are 40, 65, and 85 points, respectively, with an allowable fluctuation range of ±5 points.

[0020] Preferably, the auxiliary support force is achieved by an electrically variable damping brace, with the damping coefficient adjustable from 0.5 Nm / radian per second to 8.0 Nm / radian per second, the adjustment step size being 0.5 Nm / radian per second, and the response delay being less than 200 milliseconds.

[0021] Preferably, the voice feedback command uses natural language generation technology to generate personalized expressions based on the patient's age, educational background, and dialect preference. The speech rate is controlled at 180 to 220 words per minute, and key information is repeated twice to ensure understanding.

[0022] Preferably, the local database adopts an encrypted storage mechanism, with a data retention period of no less than 2 years. It supports uploading to the hospital rehabilitation management platform via wired or wireless means. The uploaded data is desensitized, retaining only biomechanical feature vectors and evaluation results.

[0023] Preferably, the incremental support vector machine model initiates a parameter update after completing 10 full training cycles. The input feature vector includes biomechanical scoring sequences, training load change curves, and external environmental temperature and humidity data. The output is the weight adjustment coefficient for evaluating the model in the next stage. The update process is completed within the edge computing unit and takes less than 3 minutes.

[0024] Compared with the prior art, the present invention has the following beneficial effects:

[0025] This invention achieves comprehensive quantitative monitoring of the motor function status of post-orthopedic patients by simultaneously acquiring multi-source biomechanical data, overcoming the limitations of traditional methods that focus only on a single parameter. The use of temporal alignment and joint filtering techniques ensures the synchronization and reliability of multimodal data, providing a high-quality input foundation for subsequent analysis. The constructed individualized functional assessment model transforms raw data into clinically meaningful quantitative indicators, establishing a semantic bridge between biosignatures and the rehabilitation process. The introduction of an adaptive regulation mechanism based on deviation analysis allows training intensity, frequency, and auxiliary support to be dynamically adjusted according to the patient's recovery status, forming a complete closed loop from perception and assessment to intervention. The generated medical semantic feedback instructions enhance patients' understanding and compliance with the rehabilitation process. The closed-loop learning optimization module endows the system with continuous evolution capabilities, allowing its regulation strategies to be continuously optimized with data accumulation, significantly improving the accuracy, safety, and individual fit of rehabilitation training, effectively shortening the functional recovery cycle, and reducing the risk of secondary injury due to improper training. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the overall technical solution architecture of the orthopedic postoperative rehabilitation training method based on biomechanical feedback proposed in this invention. Detailed Implementation

[0027] Example 1

[0028] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0029] Currently, with the continuous advancement of orthopedic surgical techniques, the importance of postoperative rehabilitation training in the functional recovery process of patients is becoming increasingly prominent. Modern rehabilitation medicine emphasizes individualized, dynamic, and precise training strategies to promote the coordinated repair of the skeletal, muscular, and joint systems. However, traditional rehabilitation programs often rely on standardized training manuals and physician experience, lacking an objective perception and feedback mechanism for the patient's real-time biomechanical state. This makes it difficult to precisely control the intensity, angle, and frequency of training, thus affecting rehabilitation outcomes and potentially causing secondary injuries. To address these technical problems, this invention proposes to construct an individualized biomechanical functional assessment model by simultaneously collecting multi-source biomechanical data, and to establish an adaptive adjustment mechanism for training parameters and a closed-loop learning optimization module. This achieves a complete closed-loop rehabilitation training system from perception and assessment to intervention, and is applied to a biomechanical feedback-based postoperative orthopedic rehabilitation training method.

[0030] refer to Figure 1 The schematic diagram of the overall technical architecture of the biomechanical feedback-based orthopedic postoperative rehabilitation training method proposed in this invention illustrates the logical relationships and data flow between the core components of the system. This architecture includes a multi-source biomechanical data acquisition layer, a temporal alignment and noise suppression processing layer, a personalized biomechanical functional assessment model layer, a training parameter adaptive control mechanism layer, a clinical semantic feedback instruction generation layer, and a closed-loop learning optimization module layer. The entire system interacts with patients and rehabilitation physicians through an intelligent interactive terminal and relies on a local database and edge computing units to complete data storage and model updates.

[0031] In the aforementioned biomechanical feedback-based orthopedic postoperative rehabilitation training method, step 1 involves collecting multi-source biomechanical data from the patient during rehabilitation training. This is achieved by simultaneously acquiring the three-dimensional motion angles, angular velocities, accelerations of the target joint, and the electromyographic signal intensity of related muscle groups using a wearable inertial measurement unit and a surface electromyography (EMG) sensor array. Simultaneously, a plantar pressure distribution sensor records the pressure center trajectory, stance phase duration, and bilateral load distribution ratio during the gait cycle, forming a raw data stream encompassing kinematic, dynamic, and physiological dimensions. Specifically, the wearable inertial measurement unit operates at a sampling frequency of 200 frames per second. Sensors are positioned at the distal femur, proximal tibia, and pelvic surface landmarks. A rigid body kinematics inverse kinematics algorithm is used to calculate the real-time angular changes in the three degrees of freedom of knee joint flexion / extension, adduction / abduction, and internal / external rotation, with a measurement error of less than 1.2 degrees. The surface electromyography (EMG) sensor array covers four key muscle groups: rectus femoris, vastus medialis, vastus lateralis, and semitendinosus. The electrode spacing is 20 mm, the signal gain is set to 1000 times, and the common-mode rejection ratio is greater than 110 dB. During data acquisition, the skin-electrode impedance is monitored in real time, triggering a wearing status alarm when the impedance value exceeds 5 kΩ. The plantar pressure distribution sensor is integrated into the adjustable-hardness rehabilitation insole, containing 128 piezoresistive sensing units. It has a spatial resolution of 1.5 cm x 1.5 cm, a pressure measurement range of 0 to 1200 kPa, and a sampling frequency of 100 Hz. It supports dynamic recognition of six basic activity modes, including standing, walking, and climbing stairs. All sensors transmit data to the central processing unit in real time via Bluetooth Low Energy 5.0, ensuring the continuity and integrity of the data stream.

[0032] In the aforementioned biomechanical feedback-based orthopedic postoperative rehabilitation training method, step 2 involves temporal alignment and noise suppression of the multi-source biomechanical data. Wavelet transform combined with an adaptive Kalman filter algorithm is used to correct baseline drift and filter high-frequency interference in the electromyographic (EMG) signals. Spline interpolation is applied to the kinematic data to compensate for time misalignment caused by signal transmission delay, ensuring that the synchronization accuracy of each modality data on the time axis is better than 5 milliseconds. Specifically, the EMG signals are first decomposed using a five-layer discrete wavelet transform. The db4 wavelet basis function is selected, and threshold soft contraction is applied to the detail coefficients of layers 1 to 3 to remove high-frequency noise. An adaptive Kalman filter is applied to the approximation coefficients of layers 4 to 5 for baseline drift correction. The process noise covariance Q and observation noise covariance R are dynamically adjusted according to the signal-to-noise ratio. The temporal misalignment of kinematic data mainly stems from differences in startup latency between different sensor modules and wireless transmission jitter. The system uses high-precision hardware timestamps to mark the acquisition time of each data packet and employs a cubic spline interpolation algorithm to reconstruct each modal signal on a unified time grid. The time grid interval is set to 5 milliseconds to ensure the spatiotemporal consistency of multi-source data in subsequent fusion analysis. The processed data is stored in a structured format in a memory buffer for later use by the evaluation model.

[0033] In the aforementioned biomechanical feedback-based orthopedic postoperative rehabilitation training method, step 3 involves constructing an individualized biomechanical function assessment model. Based on preset rehabilitation stage division criteria, the processed biomechanical data is mapped to four core assessment dimensions: joint range of motion recovery index, muscle activation symmetry index, gait stability index, and load tolerance growth rate. A dynamic biomechanical function score is generated through a weighted fusion algorithm, where the weights of each dimension are individually configured according to the type of surgery, postoperative days, and the patient's basic physical fitness parameters. Specifically, the joint range of motion recovery index is calculated based on the decay rate of the difference between the maximum active flexion angle of the affected and unaffected joints. The daily decay rate is set to be greater than 5% from postoperative day 14 to day 28. If the decay rate is less than 3% for three consecutive days, it is marked as delayed recovery. The muscle activation symmetry index is determined by calculating the ratio of the integrated electromyographic signals of the same muscle groups on both sides during the same movement phase. The normal range is 0.85 to 1.15. When the ratio is lower than 0.7 or higher than 1.3, a muscle synergy training program is initiated. The gait stability index is a comprehensive score based on the path length of the pressure center trajectory, the coefficient of variation of velocity, and the difference in load distribution between the left and right sides, with weights of 40%, 35%, and 25%, respectively. A balance training mode is automatically activated when the score is below 60. The load tolerance growth rate is calculated by analyzing the slope of the maximum load change that the patient can withstand during consecutive training cycles, expressed in kPa per training day. The original indicators of the four dimensions are normalized and then input into the weighted fusion module, with the following fusion formula:

[0034] S=w1·I ROM +w2·I EMG +w3·I Gait +w4·R Load

[0035] Wherein, S represents the biomechanical function score, and IROM, IEMG, IGait, and RLoad are the joint range of motion recovery index, muscle activation symmetry index, gait stability index, and load tolerance growth rate, respectively. w1, w2, w3, and w4 are the corresponding weights, satisfying w1+w2+w3+w4=1. The initial weight values ​​are preset according to the surgical type (e.g., total knee replacement, anterior cruciate ligament reconstruction) and dynamically adjusted with the number of days after surgery. For example, during the inflammation resolution period (days 1 to 14 post-surgery), w1 has a higher weight to prioritize joint range of motion recovery; during the tissue remodeling period (days 15 to 28 post-surgery), the weights of w2 and w3 are increased to enhance muscle synergy and gait symmetry; and during the functional enhancement period (from day 29 post-surgery), the weight of w4 is significantly increased to improve load tolerance. The patient's baseline physical parameters (e.g., age, body mass index, preoperative muscle strength level) are also used as weight fine-tuning factors, and correction coefficients are determined by lookup tables or linear interpolation.

[0036] In the aforementioned biomechanical feedback-based orthopedic postoperative rehabilitation training method, step 4 establishes an adaptive adjustment mechanism for training parameters. This mechanism analyzes the deviation between the biomechanical function score and the preset phased rehabilitation goals. When the score is below 90% of the target threshold, the number of repetitions of the training movements is automatically reduced to 70% of the original plan, and the auxiliary support intensity is increased by 15% to 25%. When the score is consistently above 110% of the target threshold for three consecutive assessments, the movement resistance torque is automatically increased by 10% to 20%, while the rest time between sets is shortened by 15 to 30 seconds. Specifically, the preset phased rehabilitation goals are set according to internationally accepted orthopedic rehabilitation guidelines, divided into three stages: inflammation resolution, tissue remodeling, and functional enhancement. Each stage lasts 14 days, and the corresponding biomechanical function score baselines for each stage are 40, 65, and 85 points, with an allowable fluctuation range of ±5 points. The system calculates a function score after each set of training movements and compares it with the target threshold for the current stage. If the score is below 90% of the target threshold, the system determines that the patient is in a state of delayed recovery or fatigue, and immediately triggers a protective control strategy: the number of repetitions of the next training exercise is reduced from the planned 10 to 7, and the auxiliary support is increased by 15% to 25% using an electrically adjustable variable damping brace. The damping coefficient of the electrically adjustable variable damping brace is adjustable from 0.5 Nm / radian per second to 8.0 Nm / radian per second, with an adjustment step size of 0.5 Nm / radian per second and a response delay of less than 200 milliseconds, ensuring immediate effectiveness of the support. Conversely, if the score is consistently above 110% of the target threshold for three consecutive assessments (i.e., three consecutive training sessions), the system determines that the patient has the ability to advance, and immediately triggers a challenging control strategy: the movement resistance torque is increased by 10% to 20% (e.g., by applying reverse torque through a motor), and the rest time between sets is shortened from the planned 60 seconds to 30 to 45 seconds to enhance training intensity and cardiorespiratory endurance. All regulatory decisions are recorded in the regulatory log for subsequent learning and optimization.

[0037] In the aforementioned biomechanical feedback-based orthopedic postoperative rehabilitation training method, step 5 generates feedback instructions with clinical semantic interpretation. This translates the regulatory decisions into voice and visual prompts containing medical expressions such as "the main cause of current limited joint movement is delayed quadriceps activation" or "gait asymmetry stems from insufficient arch support on the affected side," which are then pushed to the patient and rehabilitation physician in real time via an intelligent interactive terminal. Specifically, the system incorporates a medical knowledge graph, establishing a mapping relationship between abnormal patterns of biomechanical indicators and clinical diagnostic terms. For example, when the muscle activation symmetry index is below 0.7 and the integral electromyographic signal of the rectus femoris is significantly lagging behind the healthy side, the system infers that "delayed quadriceps activation" is the main cause of limited joint movement; when the difference in load distribution between the left and right sides in the gait stability index exceeds 20% and the peak pressure in the affected arch region is low, the system infers that "insufficient arch support on the affected side" is the root cause of gait asymmetry. The voice feedback commands utilize natural language generation technology, combining the patient's age, educational background, and dialect preferences to generate personalized responses. The speech rate is controlled at 180 to 220 words per minute, with key information repeated twice to ensure comprehension. Visual prompts are overlaid on the joint movement trajectory animation in the form of a color heatmap, highlighting abnormal muscle groups or pressure distribution areas, accompanied by text descriptions. All feedback content is simultaneously pushed to the rehabilitation physician's mobile terminal for remote monitoring and intervention.

[0038] In the aforementioned biomechanical feedback-based orthopedic postoperative rehabilitation training method, step 6 involves constructing a closed-loop learning optimization module. This module stores the complete data sequence of each training session, control decision records, and physician intervention annotations in a local database. An incremental support vector machine (SVM) model is used to perform pattern mining on the historical data, periodically updating the parameter mapping rules in the biomechanical functional assessment model to improve the system's accuracy in predicting individual recovery trajectories and the adaptability of control strategies. Specifically, the local database uses an AES-256 encryption storage mechanism, with a data retention period of no less than two years. It supports uploading to the hospital's rehabilitation management platform via wired or wireless means. Uploaded data undergoes anonymization, retaining only biomechanical feature vectors and assessment results. The incremental SVM model initiates a parameter update every 10 complete training cycles. The input feature vectors include the biomechanical scoring sequence, training load change curves, and external environmental temperature and humidity data. The output is the weight adjustment coefficient for the next stage of the assessment model. The model update process is completed within an edge computing unit, taking less than 3 minutes, ensuring no disruption to daily training procedures. Through continuous learning, the system can identify individual-specific recovery patterns (such as sensitivity to specific training movements, fatigue accumulation rate, etc.) and fine-tune the thresholds and weights in the assessment model accordingly to achieve truly individualized and precise rehabilitation.

[0039] To further illustrate the technical effects of this invention, a specific application example is constructed: A 58-year-old male patient underwent total knee arthroplasty on the 10th day, entering the late stage of inflammation resolution. During system initialization, his surgical type, age, body mass index, and preoperative muscle strength level were recorded. In the initial training, the wearable inertial measurement unit measured the maximum active flexion angle of the affected knee joint at 65 degrees, while the healthy side measured 120 degrees, a difference of 55 degrees; the surface electromyography sensor showed that the rectus femoris muscle activation time was delayed by 120 milliseconds compared to the healthy side; the plantar pressure distribution sensor recorded that the affected side's stance phase duration accounted for only 42% of the gait cycle. After processing in step 2, the data synchronization accuracy reached 3.2 milliseconds. Step 3 calculated the joint range of motion recovery index to be 0.45, the muscle activation symmetry index to be 0.68, and the gait stability index to be 52 points. The load tolerance growth rate was not yet enabled. The weighted fusion biomechanical function score was 38 points, lower than 90% of the target threshold of 40 points for the inflammation resolution stage (i.e., 36 points), but close to the critical value. The system determined that there was no need to reduce the training intensity, but generated a voice prompt: "Current knee flexion is limited. It is recommended to focus on engaging the quadriceps and try to squat slowly." On day 18 (day 4 of the tissue remodeling period), the system detected three consecutive scores of 72, 74, and 73, all higher than 110% of the target threshold of 65 (i.e., 71.5). Therefore, it automatically increased the resistance torque by 15%, shortened the rest time between sets from 60 seconds to 35 seconds, and pushed a prompt: "Your recovery progress is good. The training difficulty has been increased. Please maintain the correct posture." By day 35 (day 7 of the functional enhancement period), the system found through the closed-loop learning module that the patient was sensitive to changes in resistance torque but tolerated the shortening of rest time well. Therefore, in subsequent evaluation models, the weight of resistance torque was reduced and the weight of the rate of increase in load tolerance was increased, making the regulation strategy more in line with the individual characteristics.

[0040] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

[0041] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A postoperative rehabilitation training method for orthopedic surgery based on biomechanical feedback, characterized in that, The specific steps include the following: Step 1: Collect multi-source biomechanical data of the patient during rehabilitation training. Simultaneously acquire the three-dimensional motion angle, angular velocity, acceleration of the target joint and the electromyography signal intensity of the relevant muscle groups through a wearable inertial measurement unit and surface electromyography sensor array. At the same time, use a plantar pressure distribution sensor to record the pressure center trajectory, stance phase duration and bilateral load distribution ratio in the gait cycle, forming a raw data stream covering kinematic, dynamic and physiological dimensions. Step 2: Perform time alignment and noise suppression processing on the multi-source biomechanical data. Use wavelet transform combined with adaptive Kalman filtering algorithm to correct baseline drift and filter high-frequency interference in electromyography signals. Perform spline interpolation on kinematic data to compensate for time misalignment caused by signal transmission delay, and ensure that the synchronization accuracy of each modality data on the time axis is better than 5 milliseconds. Step 3: Construct an individualized biomechanical function assessment model. Based on the preset rehabilitation stage division criteria, the processed biomechanical data is mapped to four core assessment dimensions: joint range of motion recovery index, muscle activation symmetry index, gait stability index, and load tolerance growth rate. A dynamic biomechanical function score is generated through a weighted fusion algorithm, where the weight of each dimension is personalized according to the type of surgery, postoperative days, and the patient's basic physical fitness parameters. Step 4: Establish an adaptive adjustment mechanism for training parameters. Analyze the deviation between the biomechanical function score and the preset phased rehabilitation goals. When the score is below 90% of the target threshold, automatically reduce the number of repetitions of the training movements to 70% of the original plan and increase the auxiliary support intensity by 15% to 25%. When the score is consistently above 110% of the target threshold for three consecutive assessments, automatically increase the movement resistance torque by 10% to 20% and shorten the rest time between sets by 15 to 30 seconds. Step 5: Generate feedback instructions with clinical semantic interpretation, and transform the control decision into voice and visual prompts containing medical expressions such as "the main reason for the current limited joint movement is the delayed activation of the quadriceps femoris" or "gait asymmetry is due to insufficient support of the arch of the affected side", and push them to the patient and rehabilitation physician in real time through the intelligent interactive terminal. Step 6: Construct a closed-loop learning optimization module. Store the complete data sequence of each training session, the record of regulatory decisions, and the annotation of physician interventions in the local database. Use an incremental support vector machine model to perform pattern mining on the historical data and regularly update the parameter mapping rules in the biomechanical functional assessment model to improve the system's accuracy in predicting individual recovery trajectories and the adaptability of regulatory strategies.

2. The orthopedic postoperative rehabilitation training method based on biomechanical feedback according to claim 1, characterized in that: The wearable inertial measurement unit operates at a sampling frequency of 200 frames per second. Sensors are positioned at the distal femur, proximal tibia, and pelvic surface landmarks. A rigid body kinematics inverse kinematics algorithm is used to calculate real-time angular changes in the three degrees of freedom of knee joint flexion / extension, adduction / abduction, and internal / external rotation, with a measurement error of less than 1.2 degrees. The surface electromyography (EMG) sensor array covers four key muscle groups: rectus femoris, vastus medialis, vastus lateralis, and semitendinosus. The electrode spacing is 20 mm, the signal gain is set to 1000 times, and the common-mode rejection ratio is greater than 110 dB. During data acquisition, the skin-electrode impedance is monitored in real time, triggering a wearing status alarm when the impedance value exceeds 5 kΩ. The plantar pressure distribution sensor is integrated inside an adjustable-hardness rehabilitation insole, containing 128 piezoresistive sensing units with a spatial resolution of 1.5 cm x 1.5 cm, a pressure measurement range of 0 to 1200 kPa, and a sampling frequency of 100 Hz.

3. The orthopedic postoperative rehabilitation training method based on biomechanical feedback according to claim 1, characterized in that: The joint range of motion recovery index is calculated based on the decay rate of the difference between the maximum active flexion angle of the affected and healthy sides. The daily decay rate is set to be greater than 5% from day 14 to day 28 post-surgery. If the decay rate is less than 3% for three consecutive days, it is marked as delayed recovery. The muscle activation symmetry index is determined by calculating the ratio of the integrated electromyographic signals of the corresponding muscle groups on both sides during the same movement phase. The normal range is 0.85 to 1.

15. When the ratio is lower than 0.7 or higher than 1.3, a muscle synergy training program is initiated. The gait stability index is a comprehensive score based on the path length of the pressure center trajectory, the coefficient of variation of velocity, and the difference in load distribution between the left and right sides, with weights of 40%, 35%, and 25%, respectively. When the score is lower than 60, the balance training mode is automatically activated.

4. The orthopedic postoperative rehabilitation training method based on biomechanical feedback according to claim 1, characterized in that: The preset phased rehabilitation goals are set according to internationally accepted orthopedic rehabilitation guidelines and are divided into three phases: inflammation resolution phase, tissue remodeling phase, and functional enhancement phase. Each phase lasts for 14 days, and the corresponding biomechanical function score benchmarks for each phase are 40, 65, and 85 points, respectively, with an allowable fluctuation range of ±5 points.

5. The orthopedic postoperative rehabilitation training method based on biomechanical feedback according to claim 1, characterized in that: The auxiliary support force is achieved through an electrically adjustable damping brace, with a damping coefficient adjustment range of 0.5 Nm / radian / second to 8.0 Nm / radian / second, an adjustment step of 0.5 Nm / radian / second, and a response delay of less than 200 milliseconds.

6. The orthopedic postoperative rehabilitation training method based on biomechanical feedback according to claim 1, characterized in that: The voice feedback commands use natural language generation technology to generate personalized expressions based on the patient's age, educational background, and dialect preferences. The speech rate is controlled at 180 to 220 words per minute, and key information is repeated twice to ensure understanding.

7. The orthopedic postoperative rehabilitation training method based on biomechanical feedback according to claim 1, characterized in that: The local database uses an encrypted storage mechanism, with a data retention period of no less than 2 years. It supports uploading to the hospital rehabilitation management platform via wired or wireless means. The uploaded data is anonymized, retaining only biomechanical feature vectors and evaluation results.

8. The orthopedic postoperative rehabilitation training method based on biomechanical feedback according to claim 1, characterized in that: The incremental support vector machine model initiates a parameter update every 10 complete training cycles. The input feature vector includes biomechanical scoring sequences, training load change curves, and external environmental temperature and humidity data. The output is the weight adjustment coefficient for evaluating the model in the next stage. The update process is completed within the edge computing unit and takes less than 3 minutes.

Images