A Shoulder Joint Rehabilitation Evaluation Method Based on Adaptive Grey Wolf Algorithm

Abstract

The present invention relates to the technical field of rehabilitation evaluation, and discloses a shoulder joint rehabilitation evaluation method based on an adaptive grey wolf algorithm. Muscle electrical signals, muscle density and motion parameters of a patient are obtained and a grey relational degree matrix is calculated to determine the key factors for rehabilitation evaluation. The first shoulder joint rehabilitation evaluation factor is obtained based on the key factors, and then the model parameters are obtained after processing the muscle electrical signals. Then, the adaptive grey wolf algorithm is used to optimize and adjust the model parameters and obtain the second evaluation factor. Finally, the weighted sum of the first shoulder joint rehabilitation evaluation factor and the second shoulder joint rehabilitation evaluation factor is obtained to obtain the shoulder joint rehabilitation evaluation report of the patient. By setting the adaptive grey wolf algorithm, the model parameters obtained by the adaptive grey wolf algorithm are more accurate than directly performing integral operation on the muscle electrical signals to obtain a comparison sequence and preprocessing the comparison sequence to obtain the model parameters, and can better reflect the actual rehabilitation situation of the patient.

Description

Technical Field

[0001] The present invention relates to the technical field of rehabilitation evaluation, and particularly to a shoulder joint rehabilitation evaluation method based on an adaptive grey wolf algorithm. Background Art

[0002] The shoulder joint refers to the part connecting the upper limb and the trunk, including a large part of the body such as the upper part of the arm, the armpit, the front chest area, and the back area where the scapula is located. It is composed of the glenoid cavity of the scapula and the head of the humerus, also known as the glenohumeral joint. It is a typical multi-axis ball-and-socket joint, the most flexible joint in the body, and can perform three-axis movements, namely flexion and extension on the coronal axis, adduction and abduction on the sagittal axis, internal rotation, external rotation and circumduction on the vertical axis.

[0003] About 2 million new stroke patients occur in China every year. Stroke can cause varying degrees of motor function impairment in patients' limbs, and 70%-80% of stroke patients cannot live independently due to disability. The shoulder joint is one of the limb parts with a relatively high usage frequency in human daily life. The anatomical structure of the shoulder joint is relatively complex, containing a large number of muscles. After shoulder joint trauma surgery, various sequelae or complications often occur, seriously affecting the shoulder movement function. Evaluating the upper limb function can determine the degree of its dysfunction and provide a basis for clinical treatment and rehabilitation training. Clinically, accurately and objectively evaluating and understanding the motor dysfunction of the upper limb in stroke patients has become the key to formulating personalized rehabilitation treatment plans, observing treatment effects, and analyzing prognosis.

[0004] In Chinese Invention Publication No. CN108877931B, a shoulder joint rehabilitation evaluation method, device and system are proposed. First, the electromyogram signal and motion parameters of the shoulder joint to be rehabilitated are obtained; based on the electromyogram signal and motion parameters, a grey correlation degree matrix is calculated; based on the grey correlation degree matrix, the previous rehabilitation detection data and a preset grey Verhulst model, a first reference result and a second reference result are obtained; the weighted sum of the first reference result and the second reference result is calculated to obtain the rehabilitation evaluation result of the shoulder joint to be rehabilitated. The present invention can evaluate the current rehabilitation situation by collecting its own electromyogram signal and motion parameters during the motion process, improve the rehabilitation evaluation efficiency and accuracy, and provide an accurate reference basis for the rehabilitation treatment process. However, integrating the electromyogram signal to obtain a sequence to be compared; preprocessing the sequence to be compared to obtain model parameters, and inputting the model parameters into a preset grey Verhulst model to solve for an estimated sequence, and taking the estimated sequence as the second reference result is not accurate. Since the grey Verhulst model has high requirements for data quality, the process of directly substituting the model parameters into the grey Verhulst model to solve for the estimated sequence is not accurate. Therefore, the model parameters need to be further processed to make their data quality higher and ensure that the finally obtained estimated sequence is more accurate. In addition, for rehabilitation evaluation, the parameter of muscle density can be introduced to make the evaluation report more perfect. Therefore, a shoulder joint rehabilitation evaluation method based on an adaptive grey wolf algorithm is proposed. Summary of the Invention

[0005] (1) Technical Problems to be Solved

[0006] The purpose of the present invention is to solve the problem that the grey Verhulst model has high requirements for data quality, and thus the process of directly substituting the model parameters into the grey Verhulst model to solve for the estimated sequence is not accurate, and a shoulder joint rehabilitation evaluation method based on an adaptive grey wolf algorithm is proposed.

[0007] (2) Technical Solutions

[0008] The technical solutions of the present invention for solving the above technical problems are as follows:

[0009] A shoulder joint rehabilitation evaluation method based on an adaptive grey wolf algorithm includes the following steps:

[0010] S1. Obtain the electromyogram signal E m of the patient, muscle density M d and motion parameter M p and calculate to obtain a grey correlation degree matrix, and determine the key factors X for rehabilitation evaluation n , and based on the key factors X n obtain the first shoulder joint rehabilitation evaluation factor Xi , and then for the muscle electrical signal E m , after processing, the model parameters are obtained

[0011] S2. Use the adaptive grey wolf algorithm to optimize the model parameters to obtain the optimal model parameters and obtain the second shoulder joint rehabilitation evaluation factor Y j ;

[0012] S3. Calculate the weighted sum of the first shoulder joint rehabilitation evaluation factor X i and the second shoulder joint rehabilitation evaluation factor Y j to obtain the shoulder joint rehabilitation evaluation report of the patient.

[0013] Based on the above technical solutions, the present invention can also be improved as follows.

[0014] Preferably, the S1 includes:

[0015] S1.1. Obtain the muscle electrical signal E m , muscle density M d and motion parameters M p of the shoulder joint for the patient's rehabilitation evaluation, and perform denoising processing. The motion parameters M p include motion smoothness M1 and motion trajectory deviation M2;

[0016] S1.2. Based on the muscle electrical signal E m , muscle density M d and motion parameters M p , calculate the grey relational grade matrix, and based on the grey relational grade matrix, determine the key factor X n for the rehabilitation evaluation;

[0017] S1.3. Obtain the patient's previous rehabilitation test data, and compare the data of the key factor X n in S1.2 with the data of the same key factor X n in the patient's previous rehabilitation test data to obtain the first shoulder joint rehabilitation evaluation factor X i ;

[0018] S1.4. Then perform integral operation on the muscle electrical signal E m to obtain a comparison sequence, and perform preprocessing on the comparison sequence to obtain the model parameters

[0019] Preferably, the S2 includes:

[0020] S2.1. Establish a traditional grey wolf model;

[0021] S2.2. Initialize the grey wolf population using the Latin hypercube sampling method;

[0022] S2.3. Initialize the parameters of the adaptive grey wolf algorithm;

[0023] S2.4. Calculate the individual degree of each grey wolf;

[0024] S2.5. Update the parameters of the adaptive grey wolf algorithm and calculate the positions of grey wolf individuals;

[0025] S2.6. Merge and select the grey wolf population by updating the positions of grey wolf individuals;

[0026] S2.7. Determine whether the number of iterations t is greater than the maximum number of iterations, and finally output the optimal model parameters

[0027] S2.8. The optimal model parameters are input into the preset grey Verhulst model to solve and obtain the estimated sequence, and the estimated sequence is used as the second shoulder joint rehabilitation evaluation factor Y j .

[0028] Preferably, the specific process of the denoising process in S1.1 includes:

[0029] After the acquisition, the surface electromyogram signal E m is initially denoised using the Kalman filtering algorithm, the muscle density M d is initially denoised using the wavelet transform denoising method, the movement smoothness M1 is initially denoised using the moving average method, the movement trajectory deviation M2 is initially denoised using the spatial region-based filtering method, and then the surface electromyogram signal E m , the muscle density M d and the movement parameters M p are subjected to advanced denoising processing to further reduce noise interference.

[0030] Preferably, the construction process of the ConvLSTM model includes:

[0031] Obtain the data of m patients and take it out as the basic data set. Randomly divide the basic data set into a training set, a validation set, and an initial test set according to 7:2:1. Then use Keras in Python to build a ConvLSTM model and set the basic training parameters, including the learning rate, batch size, and number of iterations. Use the training set to train the ConvLSTM model to extract high-level spatio-temporal features from the time series data. Use the validation set to further verify the ConvLSTM model. Adjust the convolution kernel size and stride according to the verification effect. Finally, divide the initial test set into K subsets, use K - 1 subsets to continue the process, and the remaining one subset is used for testing. Repeat this process K times, each time selecting a different subset as the test set, and calculate the average value of the performance metrics on all test sets as the final performance metric of the ConvLSTM model.

[0032] Preferably, in S1.4, preprocess the comparison sequence to obtain the model parameters including the following steps:

[0033] Let the comparison sequence be: x (1) =(x 1 (1), x 2 (2), x 3 (3)…, x n (n));

[0034] Perform cumulative subtraction on x (1) to obtain the sequence x (0) : x (0) (k)=x (1) (k)-x (1) (k - 1), k = 2, 3, 4…n;

[0035] Perform smoothing processing on the sequence to obtain z (1) (k);

[0036]

[0037] Based on the z (1) (k), obtain the data matrix:

[0038]

[0039] Calculate the model parameters

[0040]

[0041] Preferably, in S2.3, initializing the parameters of the adaptive grey wolf algorithm includes: optimizing the initial convergence factor a initial to make it have an adaptive characteristic. The specific formula is as follows:

[0042]

[0043] Among them, a is the adaptive convergence factor, and a initial is the initial convergence factor, t is the current iteration number, and t max is the maximum iteration number, and p is the parameter controlling the decreasing speed;

[0044] Then calculate the default coefficient vector of the adaptive grey wolf algorithm and the coefficient vector The calculation formula is:

[0045]

[0046] Among them, and are random vectors in the interval [0, 1].

[0047] Preferably, the formulas for S2.5, updating the parameters of the adaptive grey wolf algorithm and calculating the positions of grey wolf individuals are as follows:

[0048]

[0049] Among them, is the position of the search agent in the next iteration, and are the optimal solution, sub-optimal solution and third-optimal solution in the current iteration respectively, a is the convergence factor, and are random vectors in the interval [0, 1].

[0050] Preferably, in the grey wolf population merging and selection in step S2.7, first merge the parent population and the offspring population into a new population P, then sort the grey wolf individuals in the new population P using the non-dominated sorting genetic algorithm, then calculate the crowding distance of the grey wolf individuals, sort them from largest to smallest according to the crowding distance, and form a new population P from the grey wolf individuals with a crowding distance greater than the sorting value s 1 , again sort the grey wolf individuals in the new population P using the non-dominated sorting genetic algorithm, then calculate the crowding distance of the grey wolf individuals, sort them from largest to smallest according to the crowding distance, and form a new population P from the grey wolf individuals with a crowding distance greater than the sorting value s 2 , repeat this step until a new population P i is formed, and when the number of grey wolf individuals in the new population P i reaches the number b of grey wolf individuals, stop the grey wolf population merging and selection.

[0051] Preferably, in step S2.8, determine whether the iteration number t is greater than the maximum iteration number t maxThe specific process is as follows. If the iteration number t is less than the maximum iteration number t max , then repeat the gray wolf population merging and selection in step S2.6, and at the same time, the iteration number t will increase by 1 until the iteration number t is greater than the maximum iteration number t max , then the final Pareto front is obtained to determine the optimal model parameters

[0052] (III) Beneficial Effects

[0053] Compared with the prior art, the technical solution of the present application has the following beneficial technical effects:

[0054] 1. By setting the adaptive gray wolf algorithm in the present invention, the adaptive convergence factor a has the adaptive characteristic. The model parameters obtained by the adaptive gray wolf algorithm are more accurate than directly performing integral operation on the surface electromyogram, obtaining a comparison sequence, and preprocessing the comparison sequence, and can better reflect the actual rehabilitation situation of the patient.

[0055] 2. By setting the Kalman filtering algorithm in the present invention, the Kalman filtering algorithm can reduce the data transmission burden and improve the processing efficiency, reduce the influence of age, race, weight, and activity level on skin impedance, and make the surface electromyogram data more in line with the actual situation.

[0056] 3. By setting the ConvLSTM model in the present invention, the ConvLSTM model is used to perform advanced denoising processing on the surface electromyogram, muscle density, and motion parameters, which can further reduce noise interference and improve the accuracy of the data.

[0057] 4. By setting the moving window method in the present invention, the process of smoothing the comparison sequence is replaced by the moving window method, and the comparison sequence can be easily adapted by adjusting parameters such as the window size, sliding step length, and judgment condition.

[0058] 5. By setting the Latin hypercube sampling method to initialize the gray wolf population in the present invention, the initial sample set is generated by uniformly stratifying and randomly selecting sample points in the multi-dimensional space, which helps to cover a wider solution space within a limited number of sampling times. Description of the Drawings

[0059] Figure 1 is the flow chart of the present invention. Detailed Embodiments

[0060] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0061] A shoulder joint rehabilitation evaluation method based on an adaptive gray wolf algorithm includes the following steps:

[0062] S1. Obtain the patient's electromyogram signal E m , muscle density M d and motion parameters M p and calculate to obtain a gray correlation degree matrix, and determine the key factors X for rehabilitation evaluation n . According to the key factors X n obtain the first shoulder joint rehabilitation evaluation factor X i . Then, after processing the electromyogram signal E m obtain the model parameters

[0063] S2. Use the adaptive gray wolf algorithm to optimize and adjust the model parameters to obtain the optimal model parameters and obtain the second shoulder joint rehabilitation evaluation factor Y j ;

[0064] S3. Calculate the weighted sum of the first shoulder joint rehabilitation evaluation factor X i and the second shoulder joint rehabilitation evaluation factor Y j to obtain the shoulder joint rehabilitation evaluation report of the patient.

[0065] The S1 includes:

[0066] S1.1. Obtain the electromyogram signal E of the shoulder joint for the patient's rehabilitation evaluation m , muscle density M d and motion parameters M p and perform denoising processing. The motion parameters M p include motion smoothness M1 and motion trajectory deviation M2;

[0067] Considering that the patient's age, race, weight, and activity level affect skin impedance, and skin impedance is related to the measurement of the electromyogram signal E m , it is necessary to perform denoising first to reduce the interference of these factors on the measurement.

[0068] S1.2. Based on the electromyogram signal E m , muscle density M d and motion parameters Mp , the grey relational grade matrix is calculated, and based on the grey relational grade matrix, the key factor X for rehabilitation evaluation is determined n ;

[0069] S1.3. Obtain the patient's previous rehabilitation test data, and compare the data of the key factor X n in S1.2 with the data of the same key factor X n in the patient's previous rehabilitation test data to obtain the first shoulder joint rehabilitation evaluation factor X i ;

[0070] S1.4. Then perform an integral operation on the muscle electrical signal E m to obtain a comparison sequence, and preprocess the comparison sequence to obtain model parameters

[0071] In S1.1, the muscle electrical signal E m , muscle density M d , and motion parameter M p of the shoulder joint for the patient's rehabilitation evaluation are obtained. The specific steps are as follows:

[0072] (1. After the patient sits quietly for 10 minutes, apply medical alcohol to the skin to be tested for cleaning. After the alcohol volatilizes, fix the test electrode on the skin, test the muscle electrical signal, and record the muscle electrical signal in the corresponding group;

[0073] (2. Use CT to scan the shoulder joint of the patient, calculate the muscle density, and record it in the corresponding group;

[0074] (3. Use a high-precision EMG sensor and a motion capture system to collect the shoulder joint motion parameters of the subject and record them in the corresponding group. The motion parameters include motion time, motion smoothness, and motion trajectory deviation.

[0075] S2 includes:

[0076] S2.1. Establish a traditional grey wolf model;

[0077] S2.2. Initialize the grey wolf population using the Latin hypercube sampling method;

[0078] S2.3. Initialize the parameters of the adaptive grey wolf algorithm;

[0079] S2.4. Calculate the individual degree of each grey wolf;

[0080] S2.5. Update the parameters of the adaptive grey wolf algorithm and calculate the position of the grey wolf individual;

[0081] S2.6. Merge and select the grey wolf population by updating the position of the grey wolf individual

[0082] S2.7. Determine whether the number of iterations \(t\) is greater than the maximum number of iterations, and finally output the best model parameters

[0083] S2.8. The best model parameters are input into the preset grey Verhulst model to solve and obtain the estimated sequence, and the estimated sequence is used as the second shoulder joint rehabilitation evaluation factor \(Y\). j 。

[0084] The specific process of the denoising process in S1.1 includes:

[0085] After the acquisition is completed, for the surface electromyogram signal \(E\) m use the Kalman filtering algorithm to perform preliminary denoising on the data, for the muscle density \(M\) d use the wavelet transform denoising method to perform preliminary denoising, use the moving average method for the motion smoothness \(M1\) to perform preliminary denoising, use the filtering method based on the spatial region for the motion trajectory deviation \(M2\) to perform preliminary denoising, and then use the ConvLSTM model for the surface electromyogram signal \(E\) m and the muscle density \(M\) d and the motion parameter \(M\) p to perform advanced denoising processing to further reduce noise interference.

[0086] The specific steps of the wavelet transform denoising method are as follows:

[0087] Wavelet transform: Perform wavelet transform on the noisy signal to obtain the wavelet coefficients at each scale.

[0088] Threshold processing: According to a certain threshold rule, perform thresholding on the wavelet coefficients at each scale to remove the wavelet coefficients controlled by noise and retain the wavelet coefficients mainly controlled by the signal. There are various selection and processing methods for the threshold, such as default threshold denoising, given threshold denoising, and forced denoising, etc.

[0089] Inverse wavelet transform: Perform inverse wavelet transform on the wavelet coefficients after threshold processing to obtain the denoised signal.

[0090] The wavelet transform denoising method has a wide range of applications in practical applications, such as in the fields of image processing, speech signal processing, biomedical signal processing, mechanical fault diagnosis, etc. In image processing, the wavelet transform denoising method can effectively remove the noise in the image and improve the clarity and quality of the image.

[0091] The Kalman filtering algorithm includes:

[0092] State prediction, covariance prediction, Kalman gain, state update, and updated state estimation error covariance;

[0093] Status prediction:

[0094] in, The prediction of the state at time k based on the information at time k-1, A is the state transfer matrix, B is the control matrix, μ k-1 is the control input at time k-1;

[0095] Covariance prediction: P k|k-1 =AP k-1|k-1 A T +Q

[0096] Among them, P k|k-1 is the covariance of the prediction errors, P k-1|k-1 is the estimation error covariance at the previous moment, Q is the covariance of the process noise;

[0097] Kalman gain: K k =P k|k-1 C T (CP k|k-1 C T +R) -1

[0098] Among them, K k is the Kalman gain, C is the observation matrix, and R is the covariance of the observation noise;

[0099] Status Update:

[0100] in, is the updated state estimate, z k is the observed value at time k;

[0101] Update state estimation error covariance: P k =(IK k H k ) k

[0102] Among them, P k is the updated state estimation error covariance matrix at the current time k, and I is the unit matrix.

[0103] The construction process of the ConvLSTM model includes:

[0104] Retrieve the data of m patients as the basic dataset. Randomly divide the basic dataset into a training set, a validation set, and an initial test set according to the ratio of 7:2:1. Then, use Keras in Python to build a ConvLSTM model and set the basic training parameters, including the learning rate, batch size, and number of iterations. Use the training set to train the ConvLSTM model to extract high-level spatio-temporal features from the time series data. Use the validation set to further validate the ConvLSTM model. Adjust the convolution kernel size and stride according to the validation effect. Finally, divide the initial test set into K subsets, use K - 1 subsets to continue the training, and the remaining one subset is used for testing. Repeat this process K times, each time selecting a different subset as the test set, and calculate the average value of the performance metrics on all test sets as the final performance metric of the ConvLSTM model.

[0105] The basic ConvLSTM model is as follows:

[0106] from keras.models import Sequential

[0107] from keras.layers import ConvLSTM2D,BatchNormalization

[0108] # Define the model

[0109] model = Sequential()

[0110] model.add(ConvLSTM2D(filters = 64, kernel_size=(3, 3), input_shape=(time_steps, rows, cols, channels), padding='same', return_sequences=True))

[0111] model.add(BatchNormalization())

[0112] model.add(ConvLSTM2D(filters = 32, kernel_size=(3, 3), padding='same', return_sequences=False))

[0113] model.add(BatchNormalization())

[0114] It should be noted that there are some syntax errors in the original code snippets you provided. For example, in line 18, "filters=64" should be "filters = 64", and in line 24, "paddi ng='same'" should be "padding='same'". The above translation has corrected these errors for better readability.model.add(Conv2D(filters=1, kernel_size=(1, 1), activation='sigmoid'))

[0115] # Compile the model

[0116] model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])。

[0117] The S1.4 preprocesses the comparison sequence to obtain model parameters including the following steps:

[0118] Let the comparison sequence be: x (1) =(x 1 (1), x 2 (2), x 3 (3)…, x n (n));

[0119] Perform cumulative subtraction on x (1) to obtain the sequence x (0) : x (0) (k) = x (1) (k) - x (1) (k - 1), k = 2, 3, 4…n;

[0120] Smooth the sequence to obtain z (1) (k);

[0121]

[0122] Based on the z (1) (k), obtain the data matrix:

[0123]

[0124] Calculate the model parameters

[0125]

[0126] The S2.3, initializing the parameters of the adaptive grey wolf algorithm includes: optimizing the initial convergence factor a initial to make it have adaptive characteristics. The specific formula is as follows:

[0127]

[0128] where a is the adaptive convergence factor, a initialis the initial convergence factor, t is the current iteration number, t max is the maximum iteration number, and p is the parameter controlling the decreasing speed;

[0129] The setting process of p is as follows:

[0130] 1. Initialization: First, set an initial value p0 for p. This value can be selected based on experience or determined by some heuristic method.

[0131] 2. Evaluate fitness: In each iteration, calculate the fitness of all candidate solutions.

[0132] 3. Calculate fitness improvement: For each candidate solution, calculate the improvement in its current fitness compared to the fitness in the previous iteration.

[0133] 4. Update p: Update p according to the fitness improvement.

[0134] Here are several methods:

[0135] Fixed decay: Regardless of the fitness improvement, decrease p by a fixed ratio or step size.

[0136] Decay based on improvement: If the fitness has a significant improvement, decrease p by a smaller amount; if the improvement is small or there is no improvement, increase the amount by which p is decreased.

[0137] Dynamic adjustment: Dynamically adjust the update rule of p according to the historical record of fitness improvement. For example, some form of moving average or exponential smoothing can be used to track the trend of fitness improvement and adjust p accordingly.

[0138] A simplified example formula is:

[0139]

[0140] where p new is the new parameter controlling the decreasing speed, p old is the old parameter controlling the decreasing speed, Δf is the fitness improvement of the current candidate solution, Δf max is the maximum fitness improvement among all candidate solutions, and α is a constant controlling the decay rate.

[0141] 5. Apply p: Use the updated p value to adjust other parameters in the algorithm (such as the exploration / exploitation coefficient) or control the search behavior.

[0142] 6. Continue iteration: Repeat steps 2 - 5 until the stopping criterion is met (such as reaching the maximum iteration number, the fitness reaching a preset threshold, or the fitness improvement being less than a certain threshold).

[0143] Recalculate the default coefficient vector of the adaptive grey wolf algorithm and the coefficient vector The calculation formula is as follows:

[0144]

[0145] where and are random vectors in the interval [0, 1].

[0146] In step S2.5, the formula for updating the parameters of the adaptive grey wolf algorithm and calculating the position of grey wolf individuals is as follows:

[0147]

[0148] where is the position of the search agent in the next iteration, and are the optimal solution, sub-optimal solution, and third-optimal solution in the current iteration respectively, a is the convergence factor, and are random vectors in the interval [0, 1].

[0149] In the grey wolf population merging and selection in step S2.7, first, the parent population and the offspring population are merged into a new population P. Then, the non-dominated sorting genetic algorithm is used to sort the grey wolf individuals in the new population P. Subsequently, the crowding distance of the grey wolf individuals is calculated and sorted from largest to smallest. The grey wolf individuals with a crowding distance greater than the sorting value s form a new population P 1 , and again, the non-dominated sorting genetic algorithm is used to sort the grey wolf individuals in the new population P. Subsequently, the crowding distance of the grey wolf individuals is calculated and sorted from largest to smallest. The grey wolf individuals with a crowding distance greater than the sorting value s form a new population P 2 , and this step is repeated until a new population P i is formed. When the number of grey wolf individuals in the new population P i reaches the number b of grey wolf individuals, the grey wolf population merging and selection stops.

[0150] In step S2.8, the specific process of judging whether the iteration number t is greater than the maximum iteration number t max is as follows. If the iteration number t is less than the maximum iteration number t max , then the grey wolf population merging and selection in step S2.6 is repeatedly executed, and at the same time, the iteration number t is incremented by 1 until the iteration number t is greater than the maximum iteration number t max , then the final Pareto front is obtained to determine the best model parameters

[0151] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0152] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A shoulder joint rehabilitation evaluation method based on an adaptive grey wolf algorithm, characterized in that: The following steps are involved: S1. Obtain the patient's muscle electrical signal E m , muscle density M d and motion parameters M p And calculate the grey correlation matrix to determine the key factors X of rehabilitation evaluation n , based on the key factor X n Get the first shoulder joint rehabilitation evaluation factor X i , and then the muscle electrical signal E m After processing, the model parameters are obtained S2, using the adaptive gray wolf algorithm to adjust the model parameters Optimize and adjust to get the best model parameters And obtain the second shoulder joint rehabilitation evaluation factor Y j ; S3. Obtaining the first shoulder joint rehabilitation evaluation factor X i and the second shoulder joint rehabilitation evaluation factor Y j The weighted sum of is used to obtain the patient's shoulder joint rehabilitation evaluation report; The S2 includes: S2.

1. Establish a traditional gray wolf model; S2.2, use Latin hypercube sampling method to initialize the gray wolf population; S2.3, initializing the adaptive grey wolf algorithm parameters; S2.

4. Calculate the individual degree of each gray wolf; S2.5, update the parameters of the adaptive gray wolf algorithm and calculate the individual positions of gray wolves; S2.

6. Merge and select the gray wolf population by updating the positions of individual gray wolves; S2.

7. Determine whether the number of iterations t is greater than the maximum number of iterations, and finally output the optimal model parameters S2.

8. Set the optimal model parameters Input the preset grey Verhulst model and solve to obtain the estimated series, which is used as the second shoulder joint rehabilitation evaluation factor Y j ; The S2.3, initializing the adaptive gray wolf algorithm parameters includes: initial convergence factor a initial Optimize it to make it adaptive. The specific formula is as follows: Where a is the adaptive convergence factor, a initial is the initial convergence factor, t is the current iteration number, t max is the maximum number of iterations, and p is the parameter that controls the decreasing speed; The process of setting p is as follows: Step 1, initialization: First, set an initial value p0 for p; Step 2: Evaluate fitness: In each iteration, calculate the fitness of all candidate solutions; Step 3: Calculate fitness improvement: For each candidate solution, calculate the improvement of its current fitness over the fitness in the previous iteration; Step 4: Update p: Update p according to the improvement of fitness; Then calculate the default coefficient vector of the adaptive gray wolf algorithm and coefficient vector The calculation formula is: in, and is a random vector in the interval [0,1].

2. A shoulder joint rehabilitation evaluation method based on the adaptive grey wolf algorithm according to claim 1, characterized in that: The S1 includes: S1.

1. Obtaining the muscle electrical signal E of the shoulder joint for patient rehabilitation evaluation m , muscle density M d and motion parameters M p And perform denoising, motion parameter M p Including motion smoothness M1, motion trajectory deviation M2; S1.2, based on the muscle electrical signal E m , muscle density M d and motion parameters M p , calculate the grey correlation matrix, and based on the grey correlation matrix, determine the key factor X of rehabilitation evaluation n ; S1.

3. Obtain the patient's last rehabilitation test data and convert the key factors X n The data of the same patient is the same as the key factor X in the previous rehabilitation test data of the same patient. n The data were compared to obtain the first shoulder joint rehabilitation evaluation factor X i ; S1.4, then the muscle electrical signal E m Perform an integral operation to obtain a comparison series, preprocess the comparison series, and obtain model parameters 3. A shoulder joint rehabilitation evaluation method based on the adaptive grey wolf algorithm according to claim 2, characterized in that: The specific process of S1.1 denoising includes: After the acquisition is completed, the muscle electrical signal E m The Kalman filter algorithm was used to perform preliminary denoising on the data and to estimate the muscle density M d The wavelet transform denoising method is used for preliminary denoising, the motion smoothness M1 is used for preliminary denoising using the sliding average method, and the motion trajectory deviation M2 is used for preliminary denoising using the spatial region-based filtering method. Then the ConvLSTM model is used to denoise the muscle electrical signal E m , muscle density M d and motion parameters M p Perform advanced denoising to further reduce noise interference.

4. A shoulder joint rehabilitation evaluation method based on the adaptive grey wolf algorithm according to claim 3, characterized in that: The construction process of the ConvLSTM model includes: The data of m patients were obtained as the basic data set, and the basic data set was randomly divided into training set, validation set and initial test set according to 7:2:

1. The ConvLSTM model was constructed using Keras in Python, and the basic training parameters were set. The basic training parameters included learning rate, batch size, and number of iterations. The ConvLSTM model was trained using the training set, and high-level spatiotemporal features were extracted from the time series data. The ConvLSTM model was further validated using the validation set. The convolution kernel size and step size were adjusted according to the validation effect. Finally, the initial test set was divided into K subsets, and K-1 subsets were used to proceed. The remaining subset was used for testing. This process was repeated K times, and a different subset was selected as the test set each time. The average value of the performance indicators on all test sets was calculated as the final performance indicator of the ConvLSTM model.

5. The shoulder joint rehabilitation evaluation method based on the adaptive grey wolf algorithm according to claim 2 is characterized in that: S1.4 preprocesses the comparison series to obtain model parameters The following steps are involved: Suppose the comparison sequence is: (1) =(x 1 (1),x 2 (2),x 3 (3)…,x n ( n )); x (1) Subtract to get the sequence x (0) :x (0) (k) = x (1) (k)-x (1) (k-1), k=2, 3, 4...n; Smoothing the series gives z (1) (k); Based on the z (1) (k), and get the data matrix: Calculating model parameters 6. A shoulder joint rehabilitation evaluation method based on an adaptive grey wolf algorithm according to claim 1, characterized in that: The formula for updating the adaptive gray wolf algorithm parameters and calculating the individual positions of gray wolves in S2.5 is as follows: in, is the position of the search agent in the next iteration, and are the optimal solution, the second-best solution, and the third-best solution in the current iteration, respectively, is the convergence factor, and is a random vector in the interval [0,1].

7. The shoulder joint rehabilitation evaluation method based on the adaptive grey wolf algorithm according to claim 1, characterized in that: In step S2.6, the gray wolf population is merged and selected. First, the parent population and the offspring population are merged into a new population P. Then, the gray wolf individuals in the new population P are sorted using a non-dominated sorting genetic algorithm. Then, the crowding distance of the gray wolf individuals is calculated, and the crowding distance is sorted from large to small. The gray wolf individuals whose crowding distance is greater than the sorting value s are formed into a new population P. 1 , the non-dominated sorting genetic algorithm is used again to sort the gray wolf individuals in the new population P, and then the crowding distance of the gray wolf individuals is calculated, and the crowding distance is sorted from large to small, and the gray wolf individuals with crowding distance greater than the sorting value s form a new population P 2 Repeat this step until a new population P is formed. i , the new population P i When the number of gray wolf individuals reaches the number of gray wolf individuals b, the merging and selection of gray wolf populations will be stopped.

8. The shoulder joint rehabilitation evaluation method based on the adaptive grey wolf algorithm according to claim 1, characterized in that: In step S2.7, it is determined whether the number of iterations t is greater than the maximum number of iterations t max The specific process is as follows: If the number of iterations t is less than the maximum number of iterations t max , then repeat the gray wolf population merging and selection in step S2.6, and the number of iterations t will increase by 1 until the number of iterations t is greater than the maximum number of iterations t max , then the final Pareto front is obtained to determine the optimal model parameters

Images