Rehabilitation method and system for hip and knee replacement based on mobile medical treatment

Abstract

The application discloses a hip-knee joint replacement postoperative rehabilitation method and system based on mobile medical treatment, relates to the technical field of medical rehabilitation, and discloses the hip-knee joint replacement postoperative rehabilitation method and system based on mobile medical treatment. The wearable sensor is used for monitoring patient motion data in real time and recognizing compensatory behavior. A training plan is dynamically adjusted in combination with multidimensional indexes, the problems of missing compensation recognition, feedback lag and plan rigidity in the traditional rehabilitation scheme are effectively solved, the rehabilitation effect can be improved, and the risk of adjacent joint injury can be reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical rehabilitation, in particular to a hip and knee replacement postoperative rehabilitation method and system based on mobile medical treatment. BACKGROUND

[0002] In the traditional hip and knee replacement postoperative rehabilitation process, patients usually need to regularly go to medical institutions for rehabilitation training under the guidance of professional medical staff. This way not only brings inconvenience to patients in terms of time and space, increases the economic burden, but also due to the limited medical staff resources, it is difficult to provide continuous, real-time personalized rehabilitation guidance to each patient. Especially for the elderly patients with difficulty in movement, frequent hospital visits increase their physical burden and psychological stress.

[0003] With the rapid development of mobile medical technology, it is possible to provide remote and real-time rehabilitation monitoring and guidance for patients with the help of wearable devices, mobile terminals and the Internet. However, the existing mobile medical rehabilitation scheme has many technical defects: first, these schemes can only record the movement data of patients simply, and cannot accurately identify the compensatory behavior of patients during rehabilitation training. Compensatory behavior refers to the fact that when patients perform rehabilitation actions, due to the limited function of the target joint, they excessively rely on adjacent joints or muscle groups to complete the action. This incorrect movement pattern not only affects the rehabilitation effect of the target joint, but also can cause overuse of adjacent joints and new injuries. The incidence of compensatory behavior will significantly increase in elderly patients due to muscle weakness and reduced proprioception. Secondly, the feedback mechanism of the existing rehabilitation scheme is seriously lagging behind, mainly relying on the analysis results of the post-training data report, and cannot provide real-time correction guidance during action execution. This lagging feedback makes it difficult for patients to correct the incorrect movement pattern in time, resulting in a significant reduction in rehabilitation effect. In addition, the existing scheme lacks the ability to dynamically adjust the training plan according to the real-time rehabilitation situation of the patient, and the adjustment of the training plan often needs to wait for the regular assessment of the doctor, making it difficult to meet the individualized rehabilitation needs of patients.

[0004] The above content is only used to assist in understanding the technical solutions of the present application, and does not represent the acknowledgement of the above content as prior art. SUMMARY

[0005] The main purpose of the present application is to provide a hip and knee replacement postoperative rehabilitation method and system based on mobile medical treatment, aiming to improve the rehabilitation effect after hip and knee replacement.

[0006] To achieve the above purpose, the present application provides a hip and knee replacement postoperative rehabilitation method based on mobile medical treatment, which comprises:

[0007] Obtaining a multi-modal kinematic data stream through wearable sensors worn on the target rehabilitation joint and its adjacent body parts of the patient;

[0008] determine a compensation index and a compensation type of the patient when performing an action of a current rehabilitation training plan based on the multi-modal kinematic data stream, and calculate a target joint effective range of motion excluding the compensation effect;

[0009] compare the compensation index with a preset first threshold and a preset second threshold, generate real-time feedback information containing visual prompt information and voice guidance information based on the comparison result and the compensation type, and output to a user interface; wherein the preset first threshold is less than the preset second threshold;

[0010] calculate a training quality score based on a plurality of the compensation indexes, the target joint effective range of motion, pain score data self-evaluated by the patient, and fatigue score data self-evaluated by the patient within a preset evaluation period;

[0011] obtain a historical sequence of the training quality score and a tolerance model previously established based on individual data of the patient, and adjust the current rehabilitation training plan based on the change trend of the historical sequence and the tolerance model, to generate an adjusted subsequent rehabilitation training plan.

[0012] In an embodiment, before the step of determining a compensation index and a compensation type of the patient when performing an action of a current rehabilitation training plan based on the multi-modal kinematic data stream, and calculating a target joint effective range of motion excluding the compensation effect, the method further comprises:

[0013] receiving an inputted surgery type and postoperative time of the patient;

[0014] based on the surgery type and the postoperative time, matching a corresponding rehabilitation action from a preset rehabilitation action database according to a preset matching rule;

[0015] generating an initial rehabilitation training plan based on the matched rehabilitation action, and taking it as the current rehabilitation training plan.

[0016] In an embodiment, the wearable sensor comprises an inertial measurement unit, a surface electromyography electrode, and a plantar pressure sensor; the multi-modal kinematic data stream comprises:

[0017] the main activity angle of the target rehabilitation joint and the compensation angle of at least one adjacent joint obtained by the inertial measurement unit;

[0018] the target muscle group electromyography signal and at least one compensation muscle group electromyography signal obtained by the surface electromyography electrode;

[0019] the center of plantar pressure movement trajectory obtained by the plantar pressure sensor worn on the plantar of the patient.

[0020] In an embodiment, the step of determining a compensation index and a compensation type of the patient when performing an action of a current rehabilitation training plan based on the multi-modal kinematic data stream and calculating a target joint effective range of motion excluding the compensation effect comprises:

[0021] acquiring the multi-modal kinematic data stream;

[0022] extracting feature data from the multi-modal kinematic data stream; the feature data includes a deviation value of the main activity angle from a preset standard reference angle, a variation of the compensation angle, an activation intensity ratio and a timing relationship of the target muscle group electromyographic signal and the compensation muscle group electromyographic signal, and a symmetry index of the foot pressure center movement trajectory;

[0023] inputting the extracted feature data into a preset compensation recognition model, calculating the compensation index by the compensation recognition model, and mapping the extracted feature data to a compensation behavior feature space of the preset compensation recognition model, determining the corresponding compensation type based on the classification boundary in the preset compensation behavior feature space;

[0024] taking the main activity angle of the target rehabilitation joint as the original range of motion;

[0025] based on the compensation type, determining a corresponding compensation effect coefficient reference value according to a mapping relationship between the preset compensation type and the compensation effect coefficient;

[0026] based on the compensation index, dynamically correcting the compensation effect coefficient reference value to generate a final compensation effect coefficient;

[0027] determining the target joint effective range of motion excluding the compensation effect based on the original range of motion and the final compensation effect coefficient.

[0028] In an embodiment, the method further comprises:

[0029] synchronously acquiring multi-modal kinematic data stream as training sample data during the execution of standard rehabilitation actions and pre-defined typical compensation actions;

[0030] annotating whether the action corresponding to the training sample data occurs compensation and the compensation type to generate annotated sample data;

[0031] training a machine learning model using the annotated sample data to obtain a trained compensation recognition model, and deploying it to obtain a preset compensation recognition model.

[0032] In an embodiment, the step of comparing the compensation index with the preset first threshold and the preset second threshold, generating real-time feedback information containing visual prompt information and voice guidance information based on the comparison result and the compensation type, and outputting to the user interface comprises:

[0033] generating a real-time action skeleton graph based on the multi-modal kinematics data stream;

[0034] superimposing the real-time action skeleton graph on the rehabilitation action guidance video played on the user interface, and marking the target rehabilitation joint and the adjacent body part where compensation is detected on the real-time action skeleton graph to generate the visual prompt information;

[0035] when the compensation index is less than the preset first threshold, generating and outputting first-level reminder voice information; when the compensation index is not less than the preset first threshold and less than the preset second threshold, generating and outputting second-level correction guidance voice information; when the compensation index is not less than the preset second threshold, generating and outputting third-level warning voice information and pausing the current training process;

[0036] when the compensation index is not less than the preset second threshold and the duration reaches the preset duration, dynamically decomposing the current rehabilitation action into multiple sub-steps or recommending alternative auxiliary training actions, and generating corresponding guidance information.

[0037] In an embodiment, the step of calculating the training quality score based on the plurality of compensation indexes, the effective range of motion of the target joint, the patient's self-evaluation pain score data, and the patient's self-evaluation fatigue score data within the preset evaluation period comprises:

[0038] averaging the plurality of compensation indexes recorded within the preset evaluation period to obtain an average compensation index;

[0039] calculating the increment of the patient's self-evaluation pain score data in the current evaluation period compared to the last evaluation period to obtain pain score increment data;

[0040] standardizing the effective range of motion of the target joint, the pain score increment data, and the fatigue score data to obtain standardized effective range of motion of the target joint, standardized pain score increment data, and standardized fatigue score data;

[0041] combining the average compensation index, the standardized effective range of motion of the target joint, the standardized pain score increment data, and the standardized fatigue score data according to a preset weight, and performing weighted summation calculation to output the training quality score.

[0042] In an embodiment, the step of acquiring the historical sequence of the training quality scores and the tolerance model previously established based on patient individual data, and adjusting the current rehabilitation training plan based on the change trend of the historical sequence and the tolerance model to generate an adjusted subsequent rehabilitation training plan comprises:

[0043] If the training quality scores of the consecutive N evaluation periods are all higher than the preset upper threshold, the subsequent rehabilitation training plan is advanced to the next difficulty stage or the training intensity is increased to generate the adjusted subsequent rehabilitation training plan.

[0044] If the training quality scores of the consecutive M evaluation periods are all lower than the preset lower threshold, the subsequent rehabilitation training plan is reduced to the previous difficulty stage or the training intensity is reduced to generate the adjusted subsequent rehabilitation training plan, and high-risk prompt information data is generated; wherein the values of N and M are determined according to the tolerance model.

[0045] In an embodiment, the method further comprises:

[0046] The high-risk prompt information data is pushed to medical staff.

[0047] The training plan adjustment instruction fed back by the medical staff is received.

[0048] Based on the training plan adjustment instruction, the adjusted subsequent rehabilitation training plan is modified to generate a new subsequent rehabilitation training plan.

[0049] In addition, in order to achieve the above-mentioned purpose, the application also provides a mobile medical-based post-hip and knee joint replacement rehabilitation system, which comprises a memory, a processor, and a mobile medical-based post-hip and knee joint replacement rehabilitation program stored on the memory and executable on the processor. The mobile medical-based post-hip and knee joint replacement rehabilitation program is configured to implement the steps of the mobile medical-based post-hip and knee joint replacement rehabilitation method.

[0050] The mobile medical-based post-hip and knee joint replacement rehabilitation method and system provided by the application can effectively solve the problems of missing compensation identification, feedback lag, and rigid plan in traditional rehabilitation schemes by real-time monitoring of patient motion data and identification of compensation behavior through wearable sensors, dynamic adjustment of training plans based on multi-dimensional indicators, and can improve rehabilitation effect and reduce the risk of adjacent joint injury. BRIEF DESCRIPTION OF DRAWINGS

[0051] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments consistent with the application and, together with the description, serve to explain the principles of the application.

[0052] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the accompanying drawings needed to be used in the embodiments or prior art description will be briefly introduced as follows. Obviously, the drawings can also provide further drawings based on these drawings for those of ordinary skill in the art without any creative effort.

[0053] Figure 1 A flowchart provided by an embodiment of the hip and knee joint replacement postoperative rehabilitation method based on mobile medical treatment of the present application;

[0054] Figure 2 A flowchart provided by another embodiment of the hip and knee joint replacement postoperative rehabilitation method based on mobile medical treatment of the present application;

[0055] Figure 3 A flowchart provided by another embodiment of the hip and knee joint replacement postoperative rehabilitation method based on mobile medical treatment of the present application; Figure 1 A detailed flowchart of step S200 in the embodiment of the present application;

[0056] Figure 4 A flowchart provided by another embodiment of the hip and knee joint replacement postoperative rehabilitation method based on mobile medical treatment of the present application;

[0057] Figure 5 A detailed flowchart of step S300 in the embodiment of the present application; Figure 1 A detailed flowchart of step S300 in the embodiment of the present application;

[0058] Figure 6 A detailed flowchart of step S400 in the embodiment of the present application; Figure 1 A detailed flowchart of step S400 in the embodiment of the present application;

[0059] Figure 7 A flowchart provided by another embodiment of the hip and knee joint replacement postoperative rehabilitation method based on mobile medical treatment of the present application;

[0060] Figure 8 A structure diagram provided by an embodiment of the hip and knee joint replacement postoperative rehabilitation system based on mobile medical treatment of the present application.

[0061] Explanation of the reference signs:

[0062] 10, memory; 20, processor.

[0063] The object implementation, functional features and advantages of the present application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0064] The technical solutions in the present application will be clearly and completely described below with reference to the drawings in the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. The components of the present application described and shown in the drawings herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the present application, but only represents selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0065] It should be understood that similar reference numerals and letters represent similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. Meanwhile, in the description of the present application, the terms "first", "second", and the like are only used for differentiation in description, and cannot be understood as indicating or implying relative importance.

[0066] In the prior art, rehabilitation after hip and knee arthroplasty usually requires patients to regularly go to medical institutions for guidance, which is time and space limited. Although the existing mobile medical scheme can record motion data, it cannot accurately identify compensatory behavior, resulting in distorted rehabilitation effect evaluation. The feedback mechanism relies on data analysis after training, which is difficult to correct incorrect actions in time, and old patients are more likely to have compensation due to muscle weakness but lack real-time intervention. In addition, the existing scheme uses a fixed training plan, which cannot dynamically adjust the intensity according to the individual rehabilitation progress.

[0067] In order to solve the above problems, it is necessary to solve the three core difficulties of insufficient compensatory behavior recognition accuracy, feedback lag and rigid training plan. First, consider how to improve the sensitivity of compensatory detection through multi-modal data fusion, capture joint angle deviation using an inertial measurement unit, and analyze muscle activation mode differences combined with electromyography. Second, a real-time feedback mechanism needs to be established, triggering voice prompts and visual markers through threshold grading, and correcting errors in real time during action execution. Finally, a dynamic adjustment model needs to be built, integrating historical training quality data and individual tolerance parameters to achieve adaptive optimization of the rehabilitation plan.

[0068] Based on this, the embodiments of the present application provide a hip and knee arthroplasty postoperative rehabilitation method based on mobile medical treatment, referring to Figure 1 , the hip and knee arthroplasty postoperative rehabilitation method based on mobile medical treatment comprises steps S100-S500, wherein:

[0069] Step S100, acquiring a multi-modal kinematic data stream through a wearable sensor worn on a target rehabilitation joint and its adjacent body parts of a patient;

[0070] Step S200, based on the multi-modal kinematic data stream, determining a compensation index and a compensation type of the patient when performing an action of the current rehabilitation training plan, and calculating a target joint effective range of motion excluding the compensation effect;

[0071] Step S300, comparing the compensation index with a preset first threshold and a preset second threshold, generating real-time feedback information containing visual prompt information and voice guidance information based on the comparison result and the compensation type, and outputting to a user interface; wherein the preset first threshold is less than the preset second threshold;

[0072] Step S400, based on a plurality of the compensation indexes in a preset evaluation period, the target joint effective range of motion, the pain score data self-evaluated by the patient, and the fatigue score data self-evaluated by the patient, a training quality score is calculated;

[0073] Step S500, obtaining a historical sequence of the training quality score and a tolerance model pre-established based on individual data of the patient, and adjusting the current rehabilitation training plan based on the change trend of the historical sequence and the tolerance model, to generate an adjusted subsequent rehabilitation training plan.

[0074] In this embodiment, the wearable sensor refers to a detection device worn on the target joint and adjacent parts of the patient, which can be implemented by a combination of an inertial measurement unit, a surface electromyography electrode and a plantar pressure sensor, for synchronous acquisition of joint angle, electromyography signal and pressure distribution data. The multi-modal kinematic data stream refers to the time series data continuously output by the above-mentioned sensors, which contains multi-dimensional information such as main activity angle, compensation angle, muscle group activation intensity and pressure center trajectory. The compensation index refers to a numerical indicator quantifying the severity of compensation behavior, which can be calculated by feature extraction and machine learning model, for evaluating the action specification. The target joint effective range of motion refers to the real joint activity range excluding the compensation effect, which is dynamically calibrated by the original angle through the correction coefficient corresponding to the compensation type. The real-time feedback information refers to the guidance content containing visual markers and hierarchical voice prompts, which triggers different levels of intervention strategies according to the comparison result of the compensation index and the preset threshold. The training quality score refers to an evaluation value of the compensation index, joint range of motion, pain and fatigue data, which is calculated by standardization processing and weighted summation, reflecting the effect of single training. The tolerance model refers to a training intensity adaptability prediction model established based on individual data of the patient such as age, surgery type and rehabilitation stage, which can be trained by a regression algorithm, for guiding the adjustment direction of the plan.

[0075] In this embodiment, the wearable sensor continuously collects joint motion angle, muscle activation state and plantar pressure distribution data when the patient performs rehabilitation actions. The system extracts features such as main activity angle deviation, compensation angle change, muscle activation intensity ratio, etc., inputs the compensation recognition model to calculate the compensation index and classifies the compensation type. According to the compensation type, select the corresponding correction coefficient to dynamically compensate the original joint range of motion, and get the true and effective range of motion. When the real-time calculated compensation index exceeds the first threshold value, the interface superimposes the error position mark and plays the correction voice; when it exceeds the second threshold value, it pauses the training and decomposes the action steps. After each evaluation period, the system integrates the average compensation index, effective range of motion and patient self-evaluation data of multiple training, and calculates the training quality score through weighted calculation. Combined with the historical score trend and the prediction result of the tolerance model, the strength level of the subsequent training plan is automatically increased or decreased.

[0076] In this embodiment, through multi-modal data fusion and dynamic correction algorithm, the compensation influence is accurately stripped, the joint function recovery level is truly reflected, and the scheme adopts threshold triggered graded real-time intervention to correct the error mode in the action execution process. And through the linkage of quantitative evaluation index and prediction model, the individualized rehabilitation intensity is adaptively optimized. In this way, the application can accurately identify and quantify the compensation behavior in rehabilitation training, provide real-time action correction guidance, and avoid the solidification of error action patterns. By dynamically calculating the effective joint range of motion, the compensation interference is eliminated, and the reliability of rehabilitation effect evaluation is improved. Based on the training quality score and the individual tolerance plan adjustment mechanism, the training intensity can be gradually increased under the premise of ensuring rehabilitation safety, and the postoperative functional recovery period is shortened.

[0077] In a feasible implementation manner, reference is made to Figure 2 Before step S200, the method further includes steps S010-S030, wherein:

[0078] Step S010, receiving the input surgery type and postoperative time of the patient;

[0079] Step S020, based on the surgery type and the postoperative time, according to a preset matching rule, matching the corresponding rehabilitation action from the preset rehabilitation action database;

[0080] Step S030, based on the matched rehabilitation action, generating an initial rehabilitation training plan, and taking it as the current rehabilitation training plan.

[0081] In this embodiment, the surgery type refers to the specific surgical classification of hip or knee replacement surgery, which can be achieved by selecting the interface to receive the joint site and prosthesis type information input by the patient, such as total hip replacement or unicompartmental knee replacement. The postoperative time refers to the number of days from the completion of the surgery to the current time, which can be automatically obtained or manually input by the date calculation module. The preset matching rule refers to the corresponding relationship between the rehabilitation action and the surgery type, postoperative time stage established according to the clinical guidelines, which can be specifically implemented by using a classification tree or a decision table based on expert experience. The rehabilitation action database refers to a collection of standard rehabilitation actions and their execution parameters recommended for different postoperative stages, which can specifically include action name, joint range of motion, training duration, and number of repetitions fields.

[0082] In this embodiment, before the patient starts rehabilitation training, the system obtains the surgery type and postoperative time data through the interactive interface, such as selecting hip replacement surgery and inputting the 7th day after surgery. The system selects the corresponding rehabilitation action items from the database according to the preset matching rule, such as matching low-intensity flexion and extension actions for 0-14 days after surgery. Based on the screening results, an initial training plan containing action sequences and intensity parameters is generated, such as 3 sets of sitting knee flexion training per day, 10 times per set. This plan is loaded as the current training scheme into the subsequent execution process, ensuring that the training content is adapted to the patient's postoperative recovery stage.

[0083] In this embodiment, the surgery details are combined with the postoperative time through the preset matching rule, which realizes the automatic generation of personalized initial plans, avoids the subjectivity and lag of manual plan development, solves the problem of mismatch between rehabilitation training plans and individualized recovery needs of patients in the prior art, ensures that the intensity and range of initial training actions strictly correspond to the joint function recovery stage, reduces the risk of secondary injury caused by excessive high-intensity training, and avoids the situation of delaying the rehabilitation process due to overly conservative plans.

[0084] In a feasible implementation, the wearable sensor includes an inertial measurement unit, a surface electromyography electrode, and a plantar pressure sensor; the multi-modal kinematic data stream includes: the main activity angle of the target rehabilitation joint and the compensation angle of at least one adjacent joint obtained by the inertial measurement unit; the electromyography signals of the target muscle group and at least one compensatory muscle group obtained by the surface electromyography electrode; and the movement trajectory of the center of pressure on the sole obtained by the plantar pressure sensor worn on the sole of the patient.

[0085] In this embodiment, the inertial measurement unit refers to a sensor device capable of measuring the three-dimensional space motion parameters of the joint, which can be implemented by a nine-axis sensor module integrating an accelerometer, a gyroscope and a magnetometer, for capturing the real-time motion angle changes of the target joint and its adjacent joints. The surface electromyography electrode refers to a sensor attached to the skin surface to collect muscle electrophysiological signals, which can be implemented by a wireless collection device with silver-silver chloride electrode pieces, for monitoring the synergistic activation state of the target muscle group and the compensatory muscle group. The plantar pressure sensor refers to a pressure distribution detection device embedded in the insole or wearable device, which can be implemented by a flexible piezoresistive sensor array, for recording the displacement trajectory of the plantar pressure center during the movement process.

[0086] In this embodiment, during the rehabilitation training after hip and knee joint replacement surgery, the inertial measurement unit is fixed to the patient's surgical joint and adjacent joints to continuously collect the main activity angle and adjacent joint compensation angle data. The surface electromyography electrode is attached to the target muscle group (such as quadriceps femoris) and the compensatory muscle group (such as the back muscle group), respectively, to synchronously record the electromyography signal intensity and timing characteristics of the two muscle groups. The plantar pressure sensor is integrated into the patient's rehabilitation training shoes to monitor the plantar pressure distribution changes during walking or standing training in real time. The three types of sensor data are transmitted to the mobile terminal through Bluetooth to form a multi-dimensional synchronous kinematic data stream, providing a comprehensive quantitative basis for subsequent compensatory behavior recognition.

[0087] In this embodiment, by fusing the three types of data of inertial measurement, electromyography signal and pressure distribution, the kinematic characteristics, muscle activation pattern and body balance state can be captured simultaneously to establish a three-dimensional recognition model of compensatory behavior. For example, during knee joint extension training, when an abnormally increased hip joint compensation angle is detected with the back muscle group activated in advance, it can be accurately determined as a trunk compensation mode, while the prior art may misjudge it as a normal action completion based on joint angle data only, thereby achieving multi-dimensional accurate recognition of compensatory behavior and solving the defect that the traditional mobile medical solution cannot distinguish between target joint active movement and compensatory movement. By synchronously collecting joint kinematics, muscle activation timing and body balance parameters, it can effectively distinguish between protective compensation caused by pain and functional compensation caused by muscle weakness, providing reliable data support for personalized rehabilitation guidance. For example, during gait training, when the plantar pressure center trajectory deviates by more than a threshold value, combined with synchronous analysis of the contralateral hip joint compensation angle, the pelvic tilt compensation mode can be accurately identified, thereby generating a targeted center of gravity adjustment guidance strategy.

[0088] In a feasible implementation manner, referring to Figure 3 , the step S200 includes steps S210-S270, wherein:

[0089] Step S210, acquiring the multi-modal kinematic data stream;

[0090] Step S220, extracting feature data from the multi-modal kinematic data stream; the feature data includes a deviation value of the main activity angle from a preset standard reference angle, a variation of the compensation angle, an activation intensity ratio and timing relationship of the target muscle group electromyographic signal and the compensatory muscle group electromyographic signal, and a symmetry index of the foot bottom pressure center movement trajectory;

[0091] Step S230, inputting the extracted feature data into a preset compensation recognition model, calculating and outputting the compensation index through the compensation recognition model; and mapping the extracted feature data to a compensation behavior feature space of the preset compensation recognition model, determining a corresponding compensation type based on a classification boundary in the preset compensation behavior feature space;

[0092] Step S240, taking the main activity angle of the target rehabilitation joint as the original range of motion;

[0093] Step S250, based on the compensation type, determining a corresponding compensation influence coefficient reference value according to a mapping relationship between the preset compensation type and the compensation influence coefficient;

[0094] Step S260, dynamically correcting the compensation influence coefficient reference value based on the compensation index, to generate a final compensation influence coefficient;

[0095] Step S270, determining the target joint effective range of motion excluding the compensation influence based on the original range of motion and the final compensation influence coefficient.

[0096] In this embodiment, the deviation value of the main activity angle from the preset standard reference angle refers to the difference between the actual activity angle of the target joint and the ideal rehabilitation motion angle, which can be calculated by comparing the joint motion trajectory collected by the inertial measurement unit with the standard motion library, and is used to quantify the accuracy of the patient's action execution. The change amount of the compensation angle refers to the abnormal activity amplitude of the adjacent joint beyond the normal assistance range during the action execution, for example, the angle increment of the knee joint when the hip joint is flexed, which is realized by synchronous monitoring of multi-joint angle sensors, and is used to identify the spatial distribution characteristics of the compensation behavior. The activation intensity ratio and timing relationship of the target muscle group and the compensation muscle group refer to the signal intensity ratio and activation sequence of the target muscle group and the compensation muscle group in the action cycle, for example, the activation timing dislocation of the hip muscle and the waist muscle in the leg lifting action, which is realized by synchronous collection of multiple muscle group signals by surface electromyography electrodes and analysis of the time sequence correlation, and is used to reveal the dynamics mechanism of muscle compensation. The symmetry index of the center of pressure movement trajectory refers to the gravity center offset degree of the bilateral pressure distribution, for example, the difference value of the lateral displacement of the pressure center of the healthy side and the affected side in the single leg support stage, which is calculated by collecting pressure distribution data by the foot pressure sensor array, and is used to evaluate the influence of the compensation behavior on the overall posture stability.

[0097] In this embodiment, during the action execution, multi-modal kinematic data streams are acquired in real time and transmitted to the processing unit. The feature data extraction module first analyzes the main activity angle data from the inertial measurement unit, such as the hip joint flexion angle, and calculates the difference value with the preset standard reference angle to generate the deviation value. At the same time, the compensation angle change amount of the adjacent joint is obtained by comparing the synchronous data of the adjacent joint angle sensors. The electromyography signals collected by the surface electromyography electrodes are filtered and processed to extract the signal intensity peaks of the target muscle group and the compensation muscle group, calculate the ratio of the two and analyze their activation timing relationship. The dynamic pressure data recorded by the foot pressure sensor is processed by the center of gravity trajectory algorithm to generate the symmetry index. These feature data are input into the pre-trained compensation recognition model, which maps the feature data through a multi-layer neural network and outputs the compensation index and type classification result. Based on the reference influence coefficient corresponding to the compensation type, such as the reference coefficient of 0.3 corresponding to the trunk tilt compensation type, the compensation index is dynamically modified in combination with the real-time calculation, for example, when the index exceeds the threshold value, the coefficient value is adjusted in the index weighting mode. Finally, the effective activity degree value excluding the compensation interference is obtained by multiplying the original activity degree and the modified influence coefficient.

[0098] In this embodiment, through multi-modal data fusion and machine learning model, fine identification of compensatory behavior is realized, for example, muscle compensation and joint compensation types can be distinguished, and the influence coefficient is dynamically corrected based on real-time index, so that the application can accurately identify various compensation behavior patterns and quantify their influence on rehabilitation effect, solving the problem of single compensation recognition dimension and rigid correction coefficient in traditional methods. By dynamically adjusting the compensation influence coefficient, the interference of compensation factors on joint range of motion can be accurately stripped, for example, only 10% of the range of motion is deducted when mild compensation is detected, and 40% is deducted when severe compensation is detected, thereby providing reliable data basis for rehabilitation training quality evaluation. The multi-modal feature fusion mechanism effectively improves the sensitivity and specificity of compensation recognition, for example, it can capture multi-dimensional compensation features such as angle abnormalities, muscle activation abnormalities and posture imbalance, avoiding single index misjudgment.

[0099] In a possible implementation, with reference to Figure 4 , the method further includes steps S610-S630, wherein:

[0100] Step S610, during the execution of standard rehabilitation actions and pre-defined typical compensatory actions, multi-modal kinematic data streams are synchronously collected as training sample data;

[0101] Step S620, whether the actions corresponding to the training sample data occur compensation and the type of compensation are labeled, generating labeled sample data;

[0102] Step S630, using the labeled sample data to train a machine learning model, obtaining a trained compensation identification model, and deploying it to obtain a pre-set compensation identification model.

[0103] In this embodiment, standard rehabilitation movements refer to standardized joint mobility training movements defined in medical rehabilitation guidelines. Specifically, basic movements such as knee flexion and leg raises, and hip abduction can be used as templates to establish kinematic data benchmarks in the absence of compensation. Typical compensatory movements refer to non-standard movement patterns adopted by patients due to limited joint function. These can be simulated by recording common incorrect postures in patients through clinical observation, such as lumbar lateral tilting to compensate for limited hip joint mobility. The multimodal kinematic data stream includes synchronous data from inertial measurement units, surface electromyography electrodes, and plantar pressure sensors. Specifically, timestamp alignment can be used to achieve multi-source data fusion, comprehensively capturing joint angles, electromyographic signals, and pressure distribution characteristics during movement. Annotated sample data refers to the manual or semi-automatic labeling of the compensatory state and type of each movement segment in the training samples. This labeling can be completed by clinicians based on video playback and biomechanical analysis results, for example, using "lumbar compensation" or "contralateral muscle overactivation" as classification labels. Machine learning models refer to classification algorithms used to identify compensatory behaviors. Specifically, random forests, support vector machines, or convolutional neural network models can be used to establish a mapping relationship between kinematic features and compensatory types through supervised learning.

[0104] In this embodiment, during the standard rehabilitation movement execution phase, wearable devices simultaneously collect data on the primary range of motion of the target joint, the compensation angle of adjacent joints, electromyographic signals, and plantar pressure, forming a baseline dataset under non-compensatory conditions. During the typical compensatory movement simulation phase, rehabilitation therapists guide patients or healthy subjects to deliberately perform incorrect movement patterns, such as over-reliance on lumbar muscles during knee flexion, while simultaneously recording multimodal data under compensatory conditions. After merging the two types of data, professionals annotate each data segment based on the movement videos and biomechanical analysis results, marking whether it contains compensatory behavior and the specific type of compensation. The annotated sample dataset is divided into training and testing sets, and input into a machine learning model for feature extraction and pattern learning. After the trained model is deployed to a mobile terminal, it can analyze patient movement data in real time and output compensation recognition results.

[0105] In this embodiment, by constructing a complete dataset containing standard movements and typical compensatory movements, and combining multi-dimensional motion feature extraction and machine learning algorithms, refined identification of compensatory behaviors is achieved. This solves the problem of insufficient accuracy in compensatory behavior identification in existing mobile healthcare solutions. By establishing a standardized data collection process and an intelligent model training mechanism, the detection sensitivity and classification accuracy of compensatory behaviors are significantly improved. This solution provides a reliable technical foundation for subsequent real-time feedback and training plan adjustments, avoiding deviations in rehabilitation outcomes caused by misjudgments of compensatory behaviors.

[0106] In one feasible implementation, refer to Figure 5, step S300 includes steps S310-S340, wherein:

[0107] Step S310, generating a real-time action skeleton map based on the multi-modal kinematic data stream;

[0108] Step S320, superimposing the real-time action skeleton map on the rehabilitation action guidance video played on the user interface, marking the target rehabilitation joint and the adjacent body part where compensation is detected on the real-time action skeleton map, and generating the visual prompt information;

[0109] Step S330, when the compensation index is less than the preset first threshold, generating and outputting first-level reminder voice information; when the compensation index is not less than the preset first threshold and less than the preset second threshold, generating and outputting second-level correction guidance voice information; when the compensation index is not less than the preset second threshold, generating and outputting third-level warning voice information and pausing the current training process;

[0110] Step S340, when the compensation index is not less than the preset second threshold and the duration reaches the preset duration, dynamically decomposing the current rehabilitation action into multiple sub-steps or recommending alternative auxiliary training actions, and generating corresponding guidance information.

[0111] In this embodiment, the real-time action skeleton map refers to a visual model of human skeletal movement trajectory generated in real time through a multi-modal kinematic data stream, which can be realized by a three-dimensional skeletal joint coordinate mapping algorithm combined with joint angle data, and is used to intuitively show the deviation of the patient's actual action from the standard action. The compensation index grading response mechanism refers to triggering different levels of voice prompts according to the threshold interval in which the index is located, which can be realized by conditional judgment logic combined with a pre-recorded multi-level voice library, for example, triggering third-level warning voice when the index exceeds 0.7. Action dynamic decomposition refers to decomposing complex rehabilitation actions into sub-steps that can be executed independently, which can be realized by a motion trajectory segmentation algorithm combined with joint range of motion threshold detection, for example, decomposing a leg lifting action into three sub-stages of hip flexion, knee extension, and holding.

[0112] In this embodiment, during the patient's execution of the training action, the inertial measurement unit continuously collects joint angle data, and the surface electromyography electrode monitors muscle activation state. These data generate a real-time action skeleton map through a skeletal modeling algorithm. The skeleton map is superimposed on the standard action guidance video, and the target joint and the compensation part are marked in different colors to form a visual contrast. When the system detects that the compensation index reaches 0.6, the second-level correction voice is triggered to guide the patient to adjust the pelvic position; if the compensation index exceeds 0.8 for 5 seconds, the training is automatically paused and the decomposed knee flexion training step animation is generated, and the use of an elastic band to assist in completing the action is recommended.

[0113] In this embodiment, real-time evaluation of action quality is achieved through multi-modal data fusion, the threshold grading mechanism established can implement differentiated intervention strategies according to the severity of compensation, and adaptive action splitting is achieved through dynamic trajectory analysis, which can generate personalized adjustment schemes for different patients' compensation modes. In this way, the real-time correction problem in the rehabilitation training process is effectively solved, and through the synergistic effect of visual superposition and multi-level voice prompts, the patient can immediately perceive the action deviation and obtain targeted guidance. The action decomposition mechanism ensures the safety of training while maintaining the continuity of the rehabilitation process, avoiding the decline in rehabilitation effect caused by the interruption of traditional schemes. Multi-modal data fusion analysis significantly improves the accuracy of compensation recognition, enabling the system to accurately distinguish between normal compensation and abnormal compensation behavior.

[0114] In a feasible implementation manner, referring to Figure 6 , step S400 comprises steps S410-S440, wherein:

[0115] Step S410, average calculation is performed on the plurality of compensation indexes recorded in the preset evaluation period to obtain an average compensation index;

[0116] Step S420, calculate the increment of the pain score data self-evaluated by the patient in the current evaluation period compared with the last evaluation period to obtain pain score increment data;

[0117] Step S430, standardization processing is performed on the target joint effective range of motion, the pain score increment data and the fatigue score data to obtain standardized target joint effective range of motion, standardized pain score increment data and standardized fatigue score data;

[0118] Step S440, the average compensation index, the standardized target joint effective range of motion, the standardized pain score increment data and the standardized fatigue score data are combined according to the preset weight and then weighted summation calculation is performed to output the training quality score.

[0119] In this embodiment, the average compensation index refers to the mean value processing of the compensation behavior quantitative indicators at multiple time points, reflecting the overall compensation degree of the patient in a training period, which can be realized by using sliding average algorithm or arithmetic average algorithm to eliminate the influence of single measurement error on the evaluation result.

[0120] The pain score increment data refers to the change in the degree of pain through the subjective feedback of the patient, and can specifically use a visual analog scale or a numerical rating scale to collect data, for evaluating the influence of rehabilitation training on the physical load of the patient. The standardization processing refers to converting indexes of different dimensions into dimensionless values of a unified dimension, and can specifically use a range standardization or Z-score standardization method to realize, for eliminating the influence of the order of magnitude difference between different indexes on the comprehensive score. The preset weight refers to the contribution ratio of each index to the training quality according to medical experience, and can specifically use an analytic hierarchy process or an expert scoring method to determine the weight distribution, for reflecting the importance difference of different indexes in the rehabilitation effect evaluation.

[0121] In the embodiment, the system automatically collects the compensation index generated by each training of the patient in the preset evaluation period, and forms a quantitative index reflecting the overall compensation level by calculating the average value of all compensation indexes in the period. At the same time, the system differentiates the pain score reported by the patient in the current period from the data of the last period to generate pain change trend data. For the three types of data of the effective range of motion of the target joint, the pain increment and the fatigue score, standardization conversion is performed respectively to eliminate the dimensional difference between the indexes. Finally, the three types of standardized data and the average compensation index are input into a weighted calculation module, and linear combination is performed according to the preset weight ratio to output a score result comprehensively reflecting the training quality.

[0122] In the embodiment, through periodic data aggregation, pain trend analysis and multi-dimensional index fusion, the dynamic evolution process of the rehabilitation state of the patient can be accurately captured, the dynamic quantitative evaluation of the training quality can be realized, the risk of misjudgment caused by single index evaluation can be avoided through the fusion of objective physiological data and subjective sensory feedback. The standardization processing ensures the comparability of data of different dimensions, and the weighted calculation model can flexibly adjust the evaluation emphasis according to the clinical needs, and provides a reliable basis for subsequent training plan adjustment. The introduction of the pain increment data can identify the critical state of the physical load of the patient in time, and prevent secondary injury caused by excessive training.

[0123] In an implementable embodiment, the step of obtaining the historical sequence of the training quality scores and the tolerance model previously established based on the individual data of the patient, and adjusting the current rehabilitation training plan based on the variation trend of the historical sequence and the tolerance model to generate an adjusted subsequent rehabilitation training plan comprises: if the training quality scores in the last N evaluation periods are all higher than the preset upper threshold, advancing the subsequent rehabilitation training plan to the next difficulty stage or increasing the training intensity to generate the adjusted subsequent rehabilitation training plan; if the training quality scores in the last M evaluation periods are all lower than the preset lower threshold, degrading the subsequent rehabilitation training plan to the previous difficulty stage or reducing the training intensity to generate the adjusted subsequent rehabilitation training plan, and generating high-risk prompt information data; wherein the values of N and M are determined according to the tolerance model.

[0124] In the embodiment, the historical sequence of the training quality scores refers to a data set of the training quality scores obtained by the patient in different evaluation periods arranged in time sequence, which can be stored and managed by a time sequence database or a structured data table, and is used to reflect the dynamic variation trend of the rehabilitation effect of the patient. The tolerance model refers to a machine learning model established based on the age, underlying disease, type of surgery and postoperative recovery data of the patient, which can be trained by using a random forest or gradient boosting algorithm, and is used to predict the adaptability of the patient to the training intensity. The values of N and M can be dynamically adjusted integer values, for example, N is 3 and M is 2, which are specifically set according to the individual adaptability parameters of the patient output by the tolerance model, and are used to balance the advancing speed and safety of the training plan.

[0125] In the embodiment, the system continuously collects the training quality scores and forms a historical sequence in the preset evaluation period. The threshold values of N and M are dynamically set by analyzing the physiological characteristics and rehabilitation progress of the individual patient through the tolerance model. When it is detected that the training quality scores continuously exceed the upper threshold, it is indicated that the current training intensity of the patient cannot meet the rehabilitation needs, and the system automatically upgrades the training plan to a higher difficulty stage, for example, increases the single training duration or introduces weight training. When the training quality scores are continuously lower than the lower threshold, the system determines that the current training intensity exceeds the bearing range of the patient, automatically reduces the training difficulty and generates a high-risk prompt, for example, replaces the deep squat training with assisted leg lifting training. During the adjustment process, the system synchronously updates the training plan parameters and generates a visual progress report.

[0126] In this embodiment, by constructing a tolerance model, combining historical training quality data, the precise matching of training intensity and patient recovery state is realized, and at the same time, through the automatic threshold judgment mechanism, timely intervention can be realized before the accumulation of compensation risk, and secondary damage can be avoided. Through the above technical scheme, the training intensity mismatch problem caused by the lag of traditional rehabilitation training plan adjustment is effectively solved. By dynamically evaluating the patient's recovery state and tolerance, the rehabilitation plan is realized on the premise of ensuring the safety of training. When the patient's recovery progress exceeds the expectation, the system automatically increases the training intensity to speed up the recovery process; when potential risks are detected, the training program is reduced in time and a warning is issued, forming a closed-loop rehabilitation management. This dynamic adjustment mechanism based on data driving significantly improves the safety and effectiveness of postoperative rehabilitation training.

[0127] In a feasible implementation manner, referring to Figure 7 , the method further comprises steps S710-S730, wherein:

[0128] Step S710, pushing the high-risk prompt information data to medical staff;

[0129] Step S720, receiving the training plan adjustment instruction fed back by the medical staff;

[0130] Step S730, modifying the adjusted subsequent rehabilitation training plan based on the training plan adjustment instruction to generate a new subsequent rehabilitation training plan.

[0131] In this embodiment, the high-risk prompt information data refers to structured data containing the number of times that the patient's current training quality score is continuously lower than the preset lower limit threshold, the pain score increment trend and the fatigue score change trend. Specifically, the patient identifier, abnormal index value and visual trend chart data package can be packaged in JSON format to realize, which is used to transmit the potential risk signal in the patient's rehabilitation process to the medical staff.

[0132] Among them, the training plan adjustment instruction refers to the rehabilitation plan modification command generated by the medical staff after analyzing the high-risk prompt information. Specifically, the instruction data containing action type adjustment and training intensity correction parameters can be transmitted by using an encrypted API interface to realize, which is used to ensure that the rehabilitation plan adjustment conforms to the clinical medical specification.

[0133] In this embodiment, when the system detects that the training quality score of the patient is lower than the preset lower threshold for a plurality of consecutive evaluation periods, a high-risk prompt information generation process is automatically triggered. The information is pushed to the mobile terminal of the responsible physician through the hospital internal medical information system, and the physician can view the recent compensatory index fluctuation, pain score change curve and fatigue accumulation trend of the patient. The physician determines whether intervention is needed to adjust the training plan according to clinical experience, for example, when it is found that the patient has muscle strain risk, the adjustment parameters of reducing the training intensity or increasing the auxiliary support action are selected through the system interface. After the adjustment instruction is returned to the system through the encryption channel, the algorithm automatically replaces the original action sequence in the training plan with the low-intensity version specified by the physician, and recalculates the daily training time allocation.

[0134] In some specific embodiments, the high-risk prompt information push can adopt a hierarchical warning mechanism, for example, when the score is lower than the threshold for 3 consecutive periods, a yellow warning is sent to the rehabilitation therapist, and when it is 5 consecutive periods, it is upgraded to a red warning and simultaneously pushed to the attending physician. The training plan adjustment instruction can include multiple modification dimensions, such as replacing the straight leg lifting action with the knee bending auxiliary leg lifting action, or adjusting the daily training frequency from 3 times to 2 times.

[0135] In this embodiment, by establishing an automatic information push channel, medical staff can intervene in time in the early stage of deterioration of compensatory behavior or aggravation of pain of the patient, avoid delay of rehabilitation progress caused by delayed intervention, realize a rapid response mechanism of medical professionals in abnormal rehabilitation training, and effectively prevent secondary injury caused by the patient's own insistence on the wrong training mode. At the same time, through the closed-loop adjustment mode of doctor-patient cooperation, it is ensured that the rehabilitation plan meets the real-time physical state of the patient and meets the clinical medical standard, significantly improving the safety and effectiveness of postoperative rehabilitation.

[0136] In the embodiments of the present application, the mobile medical-based postoperative rehabilitation method for hip and knee joint replacement can monitor patient motion data in real time through wearable sensors and identify compensatory behavior, dynamically adjust the training plan combined with multi-dimensional indicators, effectively solve the problems of missing compensatory identification, feedback lag and plan rigidity in traditional rehabilitation schemes, and can improve the rehabilitation effect and reduce the risk of adjacent joint injury.

[0137] It should be noted that the above examples are only for understanding the present application and do not constitute a limitation on the mobile medical-based postoperative rehabilitation method for hip and knee joint replacement of the present application. More forms of simple transformation based on this technical concept are within the protection scope of the present application.

[0138] The present application also provides a mobile medical-based postoperative rehabilitation system for hip and knee joint replacement, which refers to Figure 8The mobile medical rehabilitation system for hip and knee replacement surgery after the replacement surgery includes a memory 10, a processor 20, and a mobile medical rehabilitation program for hip and knee replacement surgery after the replacement surgery stored on the memory 10 and executable on the processor 20, and the mobile medical rehabilitation program for hip and knee replacement surgery after the replacement surgery is configured to implement the steps of the mobile medical rehabilitation method for hip and knee replacement surgery after the replacement surgery.

[0139] The mobile medical rehabilitation system for hip and knee replacement surgery after the replacement surgery provided by the present application adopts the mobile medical rehabilitation method for hip and knee replacement surgery after the replacement surgery in the above embodiments, and can achieve the rehabilitation effect after hip and knee replacement surgery. Compared with the prior art, the mobile medical rehabilitation system for hip and knee replacement surgery after the replacement surgery provided by the present application has the same beneficial effects as the mobile medical rehabilitation method for hip and knee replacement surgery after the replacement surgery provided by the above embodiments, and other technical features in the mobile medical rehabilitation system for hip and knee replacement surgery after the replacement surgery are the same as the features disclosed in the above method, which will not be repeated here.

[0140] It should be understood that various parts of the present application can be realized by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0141] The above is merely a specific implementation of the present application, but the protection scope of the present application is not limited thereto, and any equivalent structural transformation made by using the content of the present application specification and drawings, or direct / indirect application in other related technical fields is included in the patent protection scope of the present application.

Claims

1. A method for rehabilitation after hip and knee arthroplasty based on mobile medicine, characterized in that, The method comprises: acquiring a multi-modal kinematic data stream through a wearable sensor worn on a target rehabilitation joint of a patient and adjacent parts of the body; determining a compensation index and a compensation type of the patient when performing an action of a current rehabilitation training plan based on the multi-modal kinematic data stream, and calculating an effective range of motion of the target joint excluding the influence of compensation; comparing the compensation index with a preset first threshold and a preset second threshold, generating real-time feedback information containing visual prompt information and voice guidance information based on the comparison result and the compensation type, and outputting the real-time feedback information to a user interface; wherein the preset first threshold is less than the preset second threshold; calculating a training quality score based on a plurality of compensation indexes, the effective range of motion of the target joint, pain score data self-evaluated by the patient, and fatigue score data self-evaluated by the patient within a preset evaluation period; acquiring a historical sequence of the training quality score and a tolerance model previously established based on individual data of the patient, and adjusting the current rehabilitation training plan based on the change trend of the historical sequence and the tolerance model, to generate an adjusted subsequent rehabilitation training plan; The wearable sensor comprises an inertial measurement unit, a surface electromyography electrode, and a plantar pressure sensor; the multi-modal kinematic data stream comprises: the main activity angle of the target rehabilitation joint and the compensation angle of at least one adjacent joint acquired by the inertial measurement unit; the electromyography signals of the target muscle group and at least one compensatory muscle group acquired by the surface electromyography electrode; the moving track of the center of plantar pressure acquired by the plantar pressure sensor worn on the sole of the patient; The step of determining the compensation index and the compensation type of the patient when performing an action of a current rehabilitation training plan based on the multi-modal kinematic data stream, and calculating an effective range of motion of the target joint excluding the influence of compensation comprises: acquiring the multi-modal kinematic data stream; extracting feature data from the multi-modal kinematic data stream; the feature data comprises a deviation value of the main activity angle from a preset standard reference angle, a change amount of the compensation angle, an activation intensity ratio and a time sequence relationship of the electromyography signals of the target muscle group and the compensatory muscle group, and a symmetry index of the moving track of the center of plantar pressure; inputting the extracted feature data into a preset compensation recognition model, calculating and outputting the compensation index through the compensation recognition model; and mapping the extracted feature data to a compensation behavior feature space of the preset compensation recognition model, determining the corresponding compensation type based on the classification boundary in the preset compensation behavior feature space; taking the main activity angle of the target rehabilitation joint as the original activity degree; based on the compensation type, determining a corresponding compensation influence coefficient benchmark value according to the mapping relationship between the preset compensation type and the compensation influence coefficient; dynamically correcting the compensation influence coefficient benchmark value based on the compensation index to generate a final compensation influence coefficient; determining the effective range of motion of the target joint excluding the influence of compensation based on the original activity degree and the final compensation influence coefficient; The step of acquiring the historical sequence of the training quality scores and a tolerance model previously established based on patient individual data, and adjusting the current rehabilitation training plan based on the change trend of the historical sequence and the tolerance model to generate an adjusted subsequent rehabilitation training plan comprises: If the training quality scores of the last N evaluation periods are all higher than the preset upper threshold, the subsequent rehabilitation training plan is advanced to the next difficulty stage or the training intensity is increased to generate the adjusted subsequent rehabilitation training plan; If the training quality scores of the last M evaluation periods are all lower than the preset lower threshold, the subsequent rehabilitation training plan is reduced to the previous difficulty stage or the training intensity is reduced to generate the adjusted subsequent rehabilitation training plan, and high-risk prompt information data is generated; wherein the values of N and M are determined according to the tolerance model.

2. The mobile medical-based rehabilitation method after hip-knee arthroplasty according to claim 1, wherein, Before the step of determining the compensation index and the compensation type of the patient when performing the action of the current rehabilitation training plan based on the multi-modal kinematic data stream, and calculating the target joint effective range of motion excluding the compensation effect, the method further comprises: receiving the input surgery type and postoperative time of the patient; based on the surgery type and the postoperative time, matching the corresponding rehabilitation action from the preset rehabilitation action database according to the preset matching rule; generating an initial rehabilitation training plan based on the matched rehabilitation action as the current rehabilitation training plan.

3. The mobile medical-based rehabilitation method after hip-knee arthroplasty according to claim 1, wherein, The method further comprises: synchronously collecting multi-modal kinematic data streams as training sample data during the execution of standard rehabilitation actions and pre-defined typical compensation actions; annotating whether the action corresponding to the training sample data occurs compensation and the compensation type to generate annotated sample data; training a machine learning model using the annotated sample data to obtain a trained compensation identification model, and deploying the trained compensation identification model to obtain a preset compensation identification model.

4. The mobile medical-based rehabilitation method after hip-knee arthroplasty according to claim 1, wherein, The step of comparing the compensation index with a preset first threshold and a preset second threshold, generating real-time feedback information containing visual prompt information and voice guidance information based on the comparison result and the compensation type, and outputting the real-time feedback information to the user interface comprises: generating a real-time action skeleton graph based on the multi-modal kinematic data stream; overlaying and displaying the real-time action skeleton graph on the rehabilitation action guidance video played on the user interface, and marking the target rehabilitation joint and the adjacent body part where compensation is detected on the real-time action skeleton graph to generate the visual prompt information; when the compensation index is less than the preset first threshold, generating and outputting first-level reminder voice information; when the compensation index is not less than the preset first threshold and less than the preset second threshold, generating and outputting second-level correction guidance voice information; when the compensation index is not less than the preset second threshold, generating and outputting third-level warning voice information and pausing the current training process; when the compensation index is not less than the preset second threshold and the duration reaches the preset duration, the current rehabilitation action is dynamically decomposed into multiple sub-steps or an alternative auxiliary training action is recommended, and corresponding guidance information is generated. 5.The mobile medical-based rehabilitation method after hip and knee arthroplasty according to claim 1, wherein, The step of calculating the training quality score based on the plurality of compensation indexes, the target joint effective range of motion, the patient self-evaluated pain score data, and the patient self-evaluated fatigue score data within the preset evaluation period comprises: averaging the plurality of compensation indexes recorded within the preset evaluation period to obtain an average compensation index; calculating the increment of the patient self-evaluated pain score data in the current evaluation period compared to the last evaluation period to obtain pain score increment data; standardizing the target joint effective range of motion, the pain score increment data, and the fatigue score data to obtain standardized target joint effective range of motion, standardized pain score increment data, and standardized fatigue score data; combining the average compensation index, the standardized target joint effective range of motion, the standardized pain score increment data, and the standardized fatigue score data according to a preset weight, and then performing weighted summation calculation to output the training quality score. 6.The mobile medical-based rehabilitation method after hip and knee arthroplasty according to claim 1, wherein, The method further comprises: pushing the high-risk prompt information data to medical personnel; receiving training plan adjustment instructions fed back by medical personnel; based on the training plan adjustment instructions, modifying the adjusted subsequent rehabilitation training plan to generate a new subsequent rehabilitation training plan. 7.A mobile medical-based rehabilitation system after hip and knee arthroplasty, characterized in that, The mobile medical-based postoperative rehabilitation system for hip and knee arthroplasty comprises a memory, a processor, and a mobile medical-based postoperative rehabilitation program for hip and knee arthroplasty stored on the memory and executable on the processor, and the mobile medical-based postoperative rehabilitation program for hip and knee arthroplasty is configured to implement the steps of the mobile medical-based postoperative rehabilitation method for hip and knee arthroplasty according to any one of claims 1 to 6.

Images