Evaluation method and system for predicting rehabilitation effect of scapulohumeral periarthritis based on multiple modes

Abstract

The invention discloses an evaluation method and system for predicting a scapulohumeral periarthritis rehabilitation effect based on multiple modes, and relates to the technical field of fault detection of industrial equipment. A normal shoulder circumference database is constructed, a rehabilitation plan is loaded, initial shoulder circumference motion data are accurately collected by using a UWB and multi-modal sensor fusion technology, an initial rehabilitation period is generated by a rehabilitation prediction model according to the initial shoulder circumference motion data, and training quality and completion degree are controlled in real time through an action monitoring module. The rehabilitation trend evaluation module compares historical data with a normal value to evaluate the rehabilitation progress, and finally the model prediction optimization module performs weight calculation by integrating the rehabilitation trend and the completion degree, dynamically adjusts and generates a new rehabilitation period, and realizes personalized and precise closed-loop rehabilitation management; according to the evaluation method and system for predicting the rehabilitation effect of the scapulohumeral periarthritis based on the multiple modes, the rehabilitation training effect of the scapulohumeral periarthritis is changed from traditional experience driving to data driving, and the accuracy and flexibility of rehabilitation effect evaluation are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of fault detection of industrial equipment, in particular to a method and system for evaluating the rehabilitation effect of periarthritis based on multi-modal prediction. BACKGROUND

[0002] At present, the equipment for sports rehabilitation has been widely used in clinical and home rehabilitation training. Most of the common rehabilitation equipment relies on fixed process physical therapy, and the traditional rehabilitation method usually cannot provide real-time motion posture feedback. The evaluation of the rehabilitation effect of periarthritis depends on the subjective manual examination of doctors, the self-reported pain scale of patients or the simple joint range of motion measurement using a protractor; these methods have strong subjectivity, low evaluation frequency, cannot reflect dynamic motion quality, lack of quantitative prospective prediction and other problems.

[0003] Some existing rehabilitation equipment based on wearable sensors still mainly limit the function to action recording or real-time error correction, and the data utilization level is shallow. The potential correlation model between motion data and long-term rehabilitation effect cannot be deeply mined, and scientific and quantifiable rehabilitation progress prediction and efficacy evaluation report cannot be provided for doctors and patients. SUMMARY

[0004] The purpose of the present application is to provide a method and system for evaluating the rehabilitation effect of periarthritis based on multi-modal prediction, which can accurately collect high-dimensional motion data of patients during training by deploying a composite sensor module, predict the rehabilitation period according to the motion data, and automatically and intelligently evaluate and optimize the prediction results of single action quality, stage rehabilitation progress and final rehabilitation effect.

[0005] To achieve the above purpose, the present application provides a method for evaluating the rehabilitation effect of periarthritis based on multi-modal prediction, comprising the following steps: S1, constructing a normal periarthritis motion database and loading a rehabilitation plan; S2, collecting initial periarthritis motion data based on a UWB high-precision positioning system through a multi-modal sensor fusion module for detecting periarthritis action; S3, a rehabilitation prediction model generates an initial rehabilitation period according to the initial periarthritis motion data collected in S2, the normal periarthritis motion data in S1 and the rehabilitation plan; S4, a motion monitoring module monitors the periarthritis training data in each rehabilitation training process in real time and calculates the completion degree of the rehabilitation plan; S5, a rehabilitation trend evaluation module compares the periarthritis training data in the training process in S4 with the normal periarthritis motion data in S1 and the historical periarthritis training data to obtain the rehabilitation trend after each rehabilitation training; S6, the model prediction optimization module generates a new rehabilitation cycle according to the comprehensive weight of the rehabilitation trend of S5 and the rehabilitation plan completion of S4.

[0006] Preferably, the specific process of S1 is as follows: S11, normal shoulder joint motion data of different age groups and genders is collected, and the collected normal shoulder joint motion data includes data of shoulder joint flexion, extension, abduction, adduction, internal rotation, external rotation, horizontal adduction, horizontal abduction, and circumduction; S12, the specific rehabilitation plan given by the doctor is loaded.

[0007] Preferably, the multi-modal sensor fusion module of S2 includes a plurality of nine-axis inertial measurement units, a sensor data fusion algorithm, and a historical training database, the nine-axis inertial measurement unit includes a three-axis accelerometer, a three-axis gyroscope, and a three-axis magnetometer, the sensor data fusion algorithm includes a Kalman filter algorithm and a complementary filter algorithm; the historical training database is used to store historical shoulder joint training data of a plurality of previous times; The specific process of S2 is as follows: S21, the cumulative error of the nine-axis inertial measurement unit is corrected using a UWB high-precision positioning system, and a motion trajectory in a global coordinate system is provided; S22, the patient is guided to perform maximum attempt motion according to the shoulder joint motion in the normal shoulder joint motion database collected in S1, the position information output by the UWB is taken as an observation value, and the position information calculated by the nine-axis inertial measurement unit is fused, the sensor fusion algorithm will weight according to the noise characteristics of the two sensors to obtain relative attitude angle and kinematics parameters; S23, the sudden pause point and the point of sharp change in speed in the motion process are recorded as pain limited points; S24, the relative attitude angle and kinematics parameters obtained in S22 and the pain limited points of S23 are recorded as initial shoulder joint motion data; S25, the measured initial shoulder joint motion data and the shoulder joint training data of each training are stored in the historical database.

[0008] Preferably, the specific process of S3 is as follows: S31, the rehabilitation stage in which the patient is located is predicted according to the difference between the initial shoulder joint motion data of the patient and the normal shoulder joint motion data; S32, the initial rehabilitation cycle is generated according to the rehabilitation stage in which the patient is located and the specific rehabilitation plan given by the doctor .

[0009] Preferably, the specific process of S4 is as follows: S41, the action monitoring module monitors the action completion condition through the shoulder joint training data of the multi-modal sensor fusion module; S42, if the action is not standard and the trunk compensation problem occurs, voice prompts are given; S43, the quantity completion degree of the action in the rehabilitation plan is calculated and the quality score ; S44, the quantity completion degree and the quality score of S43 are weighted to obtain the rehabilitation plan completion degree .

[0010] Preferably, the specific calculation content of the quantity completion degree and the quality score in S43 is as follows: The quantity completion degree of a certain action The calculation formula is as follows: ; Wherein, is the actual completion times of a certain action, is the planned completion times of the corresponding same action, is the label of a certain action and , is the total number of all actions; The quality score of a certain action The calculation formula is as follows: ; Wherein, is the number of voice prompts for the action not standard and the trunk compensation problem occurring in the process of a certain action; The weight calculation formula of the rehabilitation plan completion degree in S44 is as follows: .

[0011] Preferably, the rehabilitation trend in S5 includes the average change rate and the recovery index, and the specific process of S5 is as follows: S51, according to S24, the shoulder training data in this training process and the historical shoulder training data of the previous time and the normal shoulder movement data in the normal shoulder movement database in S1 are extracted; S52, the average change rate of the shoulder training data in this time relative to the historical shoulder training data of the previous time is calculated; S53, the recovery index of the shoulder training data in this time relative to the normal shoulder movement data is calculated.

[0012] Preferably, the calculation process of the average change rate in S52 is as follows: ; Wherein, This refers to the shoulder training data collected during this training session. For the first Historical shoulder training data For training times and ; Recovery Index in S53 The calculation process is as follows: ; in, This represents normal shoulder movement data.

[0013] Preferably, the specific process of S6 is as follows: S61, Given the completion rate of the S44 rehabilitation plan S52 average rate of change and S53 recovery index The weighting coefficients are respectively , , Calculate the rehabilitation cycle adjustment coefficient using comprehensive weighting. ; Rehabilitation period adjustment coefficient The calculation formula is as follows: ; in, To the completion rate of the rehabilitation plan The weighting coefficients, Average rate of change The weighting coefficients, To recover the index weighting coefficients and ; S62, Rehabilitation cycle adjustment coefficient obtained through S61 The initial rehabilitation cycle generated by S32 Fine-tuning to obtain real-time recovery cycle ; Real-time recovery cycle The calculation formula is as follows: .

[0014] To achieve the above objectives, the present invention also provides an evaluation system for predicting the rehabilitation effect of frozen shoulder based on multimodal analysis, including a normal shoulder movement database, a multimodal sensor fusion module, a rehabilitation prediction model, a motion monitoring module, a rehabilitation trend assessment module, and a model prediction optimization module. Normal shoulder movement data from the normal shoulder movement database are input into the rehabilitation prediction model and the rehabilitation trend assessment module, respectively. The multimodal sensor fusion module collects the patient's initial shoulder movement data and inputs it into the rehabilitation prediction model, the motion monitoring module, and the rehabilitation trend assessment module, respectively. The rehabilitation prediction model predicts the initial rehabilitation cycle based on normal shoulder movement data and the patient's initial shoulder movement data, and inputs the prediction and optimization module into the model prediction module. In subsequent rehabilitation training, the motion monitoring module calculates the completion rate of the rehabilitation plan by real-time monitoring of the patient's shoulder training data collected by the multimodal sensor fusion module and inputs it into the model prediction and optimization module; The rehabilitation trend assessment module calculates the average rate of change and recovery index based on the shoulder training data, historical shoulder training data and normal shoulder movement data collected by the multimodal sensor fusion module, and inputs them into the model prediction and optimization module. The model prediction and optimization module combines the average change rate of the weights, the recovery index, and the completion rate of the rehabilitation plan to generate a rehabilitation cycle adjustment coefficient, which adjusts the initial rehabilitation cycle to obtain the real-time rehabilitation cycle.

[0015] Therefore, the present invention provides a multimodal prediction method and system for evaluating the rehabilitation effect of frozen shoulder, which, compared with the prior art, has the following advantages: 1. This application constructs a complete closed loop of initial generation—real-time monitoring—trend assessment—cycle adjustment, which can automatically fine-tune the rehabilitation cycle, supervise patients to use the correct muscle groups to exert force, ensure the effectiveness of rehabilitation training, prevent the solidification of incorrect movements, and improve the safety of training. It adopts a motion data acquisition architecture of a nine-axis inertial measurement unit and a UWB high-precision positioning system, which is specifically used for the precise capture of complex movements of the human upper limb, especially the shoulder joint. 2. The multi-sensor fusion technology solves the problem of inaccurate motion data collection in home rehabilitation scenarios, the intelligent algorithm solves the pain point of the inability to dynamically adjust rehabilitation plans, and an objective pain and movement quality assessment system is established. This has shifted the rehabilitation training effect of frozen shoulder from traditional experience-driven to data-driven, significantly improving the accuracy and flexibility of rehabilitation effect assessment.

[0016] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0017] Figure 1 This is a flowchart of an evaluation method for predicting the rehabilitation effect of frozen shoulder based on multimodal methods, according to the present invention. Figure 2 This is an overall structural diagram of an evaluation system for predicting the rehabilitation effect of frozen shoulder based on multimodal methods, according to the present invention. Detailed Implementation

[0018] In the description of this invention, it should be noted that the terms "upper," "lower," "inner," "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship in which the product of this invention is usually placed when in use. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting this invention.

[0019] Example like Figure 1 As shown, the present invention provides a method for the rehabilitation of frozen shoulder based on multimodal perception and adaptive prediction, comprising the following steps: S1. Construct a database of normal shoulder movements and load it into a rehabilitation plan; S11. Collects normal shoulder movement data of people of different ages and genders. The collected normal shoulder movement data includes data on shoulder joint flexion, extension, abduction, adduction, internal rotation, external rotation, horizontal adduction, horizontal abduction, and circumduction. S12. Load the targeted rehabilitation plan provided by the doctor; S2. The UWB-based high-precision positioning system collects initial shoulder motion data through a multimodal sensor fusion module that detects shoulder movements. The multimodal sensor fusion module includes several nine-axis inertial measurement units (IMUs), sensor data fusion algorithms, and a historical training database. Each nine-axis IMU comprises a three-axis accelerometer, a three-axis gyroscope, and a three-axis magnetometer. These IMUs are fixed to key anatomical locations on the patient's thoracic spine, scapula, upper arm, and forearm to acquire three-axis acceleration, three-axis angular velocity, and three-axis magnetic force data for each limb segment. The relative posture angles and kinematic parameters between joints are calculated using sensor fusion algorithms, including Kalman filtering and complementary filtering algorithms. The historical training database stores data from previous shoulder training sessions. S21. Use the UWB high-precision positioning system to correct the cumulative error of the nine-axis inertial measurement unit and provide the motion trajectory in the global coordinate system; The UWB high-precision positioning system includes at least three fixed base stations arranged in the training space, and several UWB tags worn by the patient to provide the absolute position coordinates of key points of the limb (such as the elbow and shoulder) in three-dimensional space. S22. Guide the patient to attempt the maximum movement according to the shoulder joint movement in the normal shoulder movement database collected in S1. Use the position information output by UWB as the observation value and fuse it with the position information calculated by the nine-axis inertial measurement unit. The sensor fusion algorithm will weight the relative attitude angle and kinematic parameters according to the noise characteristics of the two sensors. S23. Record the points where the movement suddenly stops or the points where the speed changes drastically during the exercise process and set them as pain-limiting points; S24. Record the relative posture angles and kinematic parameters obtained in S22 and the pain limitation points obtained in S23 as the initial shoulder movement data; S25. Store the initial shoulder movement data and the shoulder training data from each training session into the historical database. S3. The rehabilitation prediction model generates an initial rehabilitation cycle based on the initial shoulder movement data collected in S2, the normal shoulder movement data in S1, and the rehabilitation plan. S31. Predict the patient's rehabilitation stage based on the difference between the patient's initial shoulder movement data and normal shoulder movement data consistent with the patient's age group and gender. S32. Generate an initial rehabilitation cycle by matching the patient's rehabilitation stage with the doctor's targeted rehabilitation plan. ; The rehabilitation prediction model uses a support vector machine (SVM) to match the difference between the initial shoulder movement data and the normal shoulder movement data with the rehabilitation plan as the progress bus to determine the patient's approximate progress in the rehabilitation plan. The initial rehabilitation cycle is given by combining the rehabilitation cycle of the entire rehabilitation plan and the progress in the rehabilitation plan. The SVM performs well on high-dimensional data, has strong generalization ability, is not prone to overfitting, and is simple and reliable. S4. The motion monitoring module monitors shoulder training data in real time during each rehabilitation training session and calculates the completion rate of the rehabilitation plan. S41. The motion monitoring module monitors the completion of movements by fusing shoulder training data from a multimodal sensor fusion module. S42. Provide voice prompts if there are problems with non-standard movements or trunk compensation. S43. Calculate the number and completion rate of movements in the rehabilitation plan. and quality fraction ; Quantity of a certain action completion rate The calculation formula is as follows: ; in, The actual number of times a certain action is completed. For the corresponding The planned number of times to complete the same action A label for a certain action and , The total number of all actions; Quality score of a certain action The calculation formula is as follows: ; in, The number of voice prompts given when a certain action is not performed correctly or when there is a problem with trunk compensation. S44, Complete the quantity of S43 and quality fraction Weighting is used to obtain the completion rate of the rehabilitation plan. ; Rehabilitation plan completion rate The weight calculation formula is as follows: ; The S5 rehabilitation trend assessment module compares the shoulder training data during the S4 training process with the normal shoulder movement data and historical shoulder training data during the S1 process to obtain the rehabilitation trend after each rehabilitation training. The recovery trend includes the average rate of change and the recovery index, and the specific process is as follows: S51. Extract the shoulder training data and anterior segment data from this training process based on S24. Historical shoulder training data and normal shoulder movement data in the normal shoulder movement database in S1; S52. Calculate the shoulder training data compared to the previous one. Average rate of change of historical shoulder training data; Average rate of change The calculation process is as follows: ; in, This refers to the shoulder training data collected during this training session. For the first Historical shoulder training data For training times and ; S53. Calculate the recovery index of this shoulder training data relative to normal shoulder movement data; Recovery Index The calculation process is as follows: ; in, This represents normal shoulder movement data; S6. The model prediction and optimization module generates a new rehabilitation cycle based on the rehabilitation trend in S5 and the completion rate of the rehabilitation plan in S4, with comprehensive weights. S61, Given the completion rate of the S44 rehabilitation plan S52 average rate of change and S53 recovery index The weighting coefficients are respectively , , Calculate the rehabilitation cycle adjustment coefficient using comprehensive weighting. ; Rehabilitation period adjustment coefficient The calculation formula is as follows: ; in, To the completion rate of the rehabilitation plan The weighting coefficients, Average rate of change The weighting coefficients, To recover the index weighting coefficients and , The value is 0.5. The value is 0.3. The value is 0.2; S62, Rehabilitation cycle adjustment coefficient obtained through S61 The initial rehabilitation cycle generated by S32 Fine-tuning to obtain real-time recovery cycle ; Real-time recovery cycle The calculation formula is as follows: ; like Figure 2 As shown, the shoulder periarthritis rehabilitation system based on multimodal perception and adaptive prediction of the present invention includes a normal shoulder movement database, a multimodal sensor fusion module, a rehabilitation prediction model, a motion monitoring module, a rehabilitation trend assessment module, and a model prediction optimization module. Normal shoulder movement data from the normal shoulder movement database are input into the rehabilitation prediction model and the rehabilitation trend assessment module, respectively. The multimodal sensor fusion module collects the patient's initial shoulder movement data and inputs it into the rehabilitation prediction model, the motion monitoring module, and the rehabilitation trend assessment module, respectively. The rehabilitation prediction model predicts the initial rehabilitation cycle based on normal shoulder movement data and the patient's initial shoulder movement data, and inputs the prediction and optimization module into the model prediction module. In subsequent rehabilitation training, the motion monitoring module calculates the completion rate of the rehabilitation plan by real-time monitoring of the patient's shoulder training data collected by the multimodal sensor fusion module and inputs it into the model prediction and optimization module; The rehabilitation trend assessment module calculates the average rate of change and recovery index based on the shoulder training data, historical shoulder training data and normal shoulder movement data collected by the multimodal sensor fusion module, and inputs them into the model prediction and optimization module. The model prediction and optimization module combines the average change rate of the weights, the recovery index, and the completion rate of the rehabilitation plan to generate a rehabilitation cycle adjustment coefficient, which adjusts the initial rehabilitation cycle to obtain the real-time rehabilitation cycle.

[0020] Therefore, the present invention adopts a multimodal prediction method and system for evaluating the rehabilitation effect of frozen shoulder based on the above content. By deploying a composite sensor module, it accurately collects high-dimensional motion data of patients during training, predicts the rehabilitation cycle based on the motion data, and optimizes the results through automated and intelligent evaluation and prediction of single movement quality, phased rehabilitation progress and final rehabilitation effect.

[0021] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for evaluating the rehabilitation effect of frozen shoulder based on multimodal prediction, characterized in that, Includes the following steps: S1. Construct a database of normal shoulder movements and load it into a rehabilitation plan; S2. The UWB-based high-precision positioning system collects initial shoulder motion data through a multimodal sensor fusion module that detects shoulder movements. S3. The rehabilitation prediction model generates an initial rehabilitation cycle based on the initial shoulder movement data collected in S2, the normal shoulder movement data in S1, and the rehabilitation plan. S4. The motion monitoring module monitors shoulder training data in real time during each rehabilitation training session and calculates the completion rate of the rehabilitation plan. The S5 rehabilitation trend assessment module compares the shoulder training data during the S4 training process with the normal shoulder movement data and historical shoulder training data during the S1 process to obtain the rehabilitation trend after each rehabilitation training. S6. The model prediction and optimization module generates a new rehabilitation cycle by comprehensively weighting the rehabilitation trend in S5 and the completion rate of the rehabilitation plan in S4.

2. The evaluation method for predicting the rehabilitation effect of frozen shoulder based on multimodal prediction as described in claim 1, characterized in that: The specific process of S1 is as follows: S11. Collects normal shoulder movement data of people of different ages and genders. The collected normal shoulder movement data includes data on shoulder joint flexion, extension, abduction, adduction, internal rotation, external rotation, horizontal adduction, horizontal abduction, and circumduction. S12. Load the targeted rehabilitation plan provided by the doctor.

3. The evaluation method for assessing the rehabilitation effect of frozen shoulder based on multimodal prediction as described in claim 2, characterized in that: The S2 multimodal sensor fusion module includes several nine-axis inertial measurement units, sensor data fusion algorithms, and a historical training database. The nine-axis inertial measurement units include a three-axis accelerometer, a three-axis gyroscope, and a three-axis magnetometer. The sensor data fusion algorithms include Kalman filtering and complementary filtering algorithms. The historical training database is used to store historical shoulder training data from previous training sessions. The specific process of S2 is as follows: S21. Use the UWB high-precision positioning system to correct the cumulative error of the nine-axis inertial measurement unit and provide the motion trajectory in the global coordinate system; S22. Guide the patient to attempt the maximum movement according to the shoulder joint movement in the normal shoulder movement database collected in S1. Use the position information output by UWB as the observation value and fuse it with the position information calculated by the nine-axis inertial measurement unit. The sensor fusion algorithm will weight the relative attitude angle and kinematic parameters according to the noise characteristics of the two sensors. S23. Record the points where the movement suddenly stops or the points where the speed changes drastically during the exercise process and set them as pain-limiting points; S24. Record the relative posture angles and kinematic parameters obtained in S22 and the pain limitation points obtained in S23 as the initial shoulder movement data; S25. Store the initial shoulder movement data and the shoulder training data from each training session into the historical database.

4. The evaluation method for predicting the rehabilitation effect of frozen shoulder based on multimodal prediction as described in claim 3, characterized in that: The specific process of S3 is as follows: S31. Predict the patient's rehabilitation stage based on the difference between the patient's initial shoulder movement data and normal shoulder movement data; S32. Generate an initial rehabilitation cycle based on the patient's current stage of rehabilitation and the targeted rehabilitation plan provided by the doctor. .

5. The evaluation method for predicting the rehabilitation effect of frozen shoulder based on multimodal analysis as described in claim 4, characterized in that: The specific process of S4 is as follows: S41. The motion monitoring module monitors the completion of movements by fusing shoulder training data from a multimodal sensor fusion module. S42. Provide voice prompts if there are problems with non-standard movements or trunk compensation. S43. Calculate the number and completion rate of movements in the rehabilitation plan. and quality fraction ; S44, Complete the quantity of S43 and quality fraction Weighting is used to obtain the completion rate of the rehabilitation plan. .

6. The evaluation method for predicting the rehabilitation effect of frozen shoulder based on multimodal prediction as described in claim 5, characterized in that: The specific calculations for quantity completion and quality score in S43 are as follows: Quantity of a certain action completion rate The calculation formula is as follows: ； in, The actual number of times a certain action is completed. For the corresponding The planned number of times to complete the same action A label for a certain action and , The total number of all actions; Quality score of a certain action The calculation formula is as follows: ； in, The number of voice prompts given when a certain action is not performed correctly or when there is a problem with trunk compensation. S44 Rehabilitation Plan Completion Rate The weight calculation formula is as follows: 。 7. The evaluation method for predicting the rehabilitation effect of frozen shoulder based on multimodal prediction as described in claim 6, characterized in that: The recovery trend in S5 includes the average rate of change and the recovery index. The specific process of S5 is as follows: S51. Extract the shoulder training data and anterior segment data from this training process based on S24. Historical shoulder training data and normal shoulder movement data in the normal shoulder movement database in S1; S52. Calculate the shoulder training data compared to the previous one. Average rate of change of historical shoulder training data; S53. Calculate the recovery index of this shoulder training data relative to normal shoulder movement data.

8. The evaluation method for predicting the rehabilitation effect of frozen shoulder based on multimodal prediction as described in claim 7, characterized in that: Average rate of change in S52 The calculation process is as follows: ； in, This refers to the shoulder training data collected during this training session. For the first Historical shoulder training data For training times and ; Recovery Index in S53 The calculation process is as follows: ； in, This represents normal shoulder movement data.

9. The evaluation method for predicting the rehabilitation effect of frozen shoulder based on multimodal prediction as described in claim 8, characterized in that: The specific process of S6 is as follows: S61, Given the completion rate of the S44 rehabilitation plan S52 average rate of change and S53 recovery index The weighting coefficients are respectively , , Calculate the rehabilitation cycle adjustment coefficient using comprehensive weighting. ; Rehabilitation period adjustment coefficient The calculation formula is as follows: ； in, To the completion rate of the rehabilitation plan The weighting coefficients, Average rate of change The weighting coefficients, To recover the index weighting coefficients and ; S62, Rehabilitation cycle adjustment coefficient obtained through S61 The initial rehabilitation cycle generated by S32 Fine-tuning to obtain real-time recovery cycle ; Real-time recovery cycle The calculation formula is as follows: 。 10. A multimodal prediction system for evaluating the rehabilitation effect of frozen shoulder, characterized in that: The system, which uses any one of the multimodal prediction methods for evaluating the rehabilitation effect of frozen shoulder as described in any of claims 1-9, includes a normal shoulder movement database, a multimodal sensor fusion module, a rehabilitation prediction model, a motion monitoring module, a rehabilitation trend assessment module, and a model prediction optimization module. Normal shoulder movement data from the normal shoulder movement database are input into the rehabilitation prediction model and the rehabilitation trend assessment module, respectively. The multimodal sensor fusion module collects the patient's initial shoulder movement data and inputs it into the rehabilitation prediction model, the motion monitoring module, and the rehabilitation trend assessment module, respectively. The rehabilitation prediction model predicts the initial rehabilitation cycle based on normal shoulder movement data and the patient's initial shoulder movement data, and inputs the prediction and optimization module into the model prediction module. In subsequent rehabilitation training, the motion monitoring module calculates the completion rate of the rehabilitation plan by real-time monitoring of the patient's shoulder training data collected by the multimodal sensor fusion module and inputs it into the model prediction and optimization module; The rehabilitation trend assessment module calculates the average rate of change and recovery index based on the shoulder training data, historical shoulder training data and normal shoulder movement data collected by the multimodal sensor fusion module, and inputs them into the model prediction and optimization module. The model prediction and optimization module combines the average change rate of the weights, the recovery index, and the completion rate of the rehabilitation plan to generate a rehabilitation cycle adjustment coefficient, which adjusts the initial rehabilitation cycle to obtain the real-time rehabilitation cycle.

Images