Paravertebral muscle function change and scoliosis internal correlation modeling method

By using adaptive wavelet packet transform and multimodal dynamic graph convolution model, combined with muscle-spinal stiffness differential equation, the problem of the inability to quantitatively assess the correlation between paraspinal muscle function and scoliosis in existing technologies is solved, realizing dynamic assessment and real-time correction guidance, and improving the accuracy of Cobb angle prediction.

CN120878266APending Publication Date: 2025-10-31SHENZHEN LONGGANG DISTRICT EIGHTH PEOPLES HOSPITAL
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
CN202511048542.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-07-29
Publication Date
2025-10-31

AI Technical Summary

Technical Problem

Existing technologies cannot effectively quantify the intrinsic relationship between paraspinal muscle function and scoliosis, and traditional methods cannot capture dynamic muscle synergy patterns and spinal biomechanical responses while reducing radiation risks, thus lacking closed-loop guidance.

Method used

An adaptive wavelet packet transform fused with empirical mode decomposition was used to process surface electromyography signals, and a multimodal dynamic graph convolution model was constructed. Combined with the muscle-spinal stiffness joint differential equation, the model was trained through a physical constraint loss function to realize the intrinsic correlation model between muscle function changes and scoliosis.

Benefits of technology

It enables dynamic assessment of changes in paraspinal muscle function and scoliosis, reduces radiation risk, provides real-time biofeedback and personalized correction guidance, and significantly improves the accuracy of Cobb angle prediction.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120878266A_ABST
    Figure CN120878266A_ABST
Patent Text Reader

Abstract

According to the paravertebral muscle function change and scoliosis intrinsic correlation modeling method provided by the invention, a proposed phase-dependent bilateral muscle asymmetry index brings a spine rotation phase into myoelectricity asymmetry evaluation for the first time, and the problem that a traditional static index cannot capture dynamic compensation is solved; a multi-modal dynamic graph convolution model fuses a learnable topological structure and biomechanical constraints, data driving advantages are reserved, meanwhile, a prediction result conforms to a spine mechanical law, and the prediction error of the model on a Cobb angle is reduced to be extremely low; finally, based on a real-time feedback system of a muscle-spine differential equation, BAI can be converted into a vibration prompt signal to guide the patient to correct the action, and the method is obviously superior to a traditional training method; and a first complete-cycle solution which runs through'pathological mechanism research-quantitative evaluation-real-time intervention 'is provided for the AIS.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of AIS detection and correction technology, and in particular to a method for modeling the intrinsic correlation between paravertebral muscle function changes and scoliosis. Background Technology

[0002] Adolescent idiopathic scoliosis (AIS), the most common three-dimensional spinal deformity, has a global incidence of 2-3%. Its pathological mechanism involves the interaction of multiple factors, including vertebral structural abnormalities, neuromuscular control dysregulation, and biomechanical imbalance. Currently, clinical diagnosis mainly relies on static X-ray imaging to measure the Cobb angle, but this method has three major limitations: First, ionizing radiation exposure limits the feasibility of frequent monitoring and makes it difficult to capture the dynamic progression of the disease; second, existing assessment systems neglect the intrinsic correlation between paraspinal muscle function and spinal deformity. Studies have shown that the transformation of paraspinal muscle fiber type (from type I to type II) and the asymmetry of electromyographic activity on the convex side in AIS patients can be more than three times that of healthy individuals, but quantitative assessment tools are lacking; third, traditional surface electromyography (sEMG) analysis mostly uses static postural measurements and cannot reflect abnormal muscle coordination patterns in daily life activities.

[0003] Current technologies attempt to address these issues. For example, patent document CN120093221A proposes a multimodal assessment scheme that integrates plantar pressure, sEMG, and ultrasound. However, it has significant shortcomings: (1) Feature engineering relies on artificially designed indicators (such as median electromyography frequency and foot pressure symmetry coefficient), failing to uncover deeper spatiotemporal patterns; (2) The model lacks biomechanical interpretability, and the causal inference is lost due to the fusion of features by weighting; (3) A closed-loop correlation between dynamic changes in muscle function and spinal biomechanical response has not been established. In addition, existing wearable devices (such as surface electromyography garments) are mostly limited to data acquisition and have not achieved closed-loop guidance of "assessment-prediction-intervention".

[0004] Therefore, there is an urgent need to develop a dynamic assessment method based on a muscle function quantification model, which can reduce radiation risk, reveal the pathological link between paraspinal muscles and scoliosis, and provide real-time biofeedback for personalized movement correction. Summary of the Invention

[0005] The purpose of this invention is to provide a method for modeling the intrinsic correlation between changes in paraspinal muscle function and scoliosis, so as to solve the problems existing in the prior art.

[0006] To achieve the above objectives, the present invention provides the following solution: This invention provides a method for modeling the intrinsic correlation between changes in paravertebral muscle function and scoliosis, comprising the following steps: S1. Collect surface electromyography signals and synchronous spinal kinematic data of AIS patients while they are standing and bending over. S2. Preprocess the surface electromyography signal by adaptive wavelet packet transform fusion with empirical mode decomposition, using the following formula: ; ; in, Let k be the kth order eigenmode function. This is an empirical mode decomposition operator used to decompose non-stationary signals into intrinsic mode functions. This represents the original surface electromyography signal. For the preprocessing results, For wavelet packet transform, These are Symlet-5 wavelet basis functions. For adaptive threshold, , Let N be the standard deviation of the k-th intrinsic mode function, N be the signal length, and K be the total number of effective IMF components. S3. Extract time-domain features, frequency-domain features, and spatial features from the preprocessed data; S4. Calculate the phase-dependent bilateral muscle asymmetry index based on the three-dimensional features, using the following formula: ; in, The phase angle of spinal rotation. , These are the W-RMS values ​​for the left and right channels. This is the time decay factor, with a value of 0.15. To prevent division by zero constant, a value of 0.01 is used; S5. Construct a multimodal dynamic graph convolution model; S6. Establish the joint differential equation for muscle-spine stiffness, the equation is as follows: ; in, This is the muscle stiffness coefficient. Let be the ligament stiffness constant. For the angle of scoliosis, This represents the angle of the spine in a neutral position. It is the viscous damping coefficient; S7. Train the model using the physical constraint loss function, the formula is: ; in, , These are the weighting coefficients. To predict the Cobb angle for the model, To measure the Cobb angle by X-ray, This is the right-hand side term of the differential equation.

[0007] Preferably, in step S1, the surface electromyography signal includes 64 channels of data.

[0008] Preferably, in step S1, the kinematic data includes the spinal rotation phase angle τ and the Cobb angle, and the kinematic data is acquired by an inertial measurement unit.

[0009] Preferably, step S2 further includes motion artifact elimination, with the following formula: ; in, The vertical motion coupling coefficient is... The horizontal damping factor is... Vertical acceleration, This represents the magnitude of the horizontal plane acceleration vector.

[0010] Preferably, in step S3, the time-domain feature is an improved weighted root mean square value, and the formula is: ; ; in, Let i be the electromyographic amplitude at the i-th sampling point. The time window weighting function, The number of sampling points in the analysis window; The frequency domain feature is the electromyographic power spectrum entropy, and the formula is: ; in, For the normalized power spectrum, The entropy of the electromyographic power spectrum; The spatial feature is the paraspinal muscle activation gradient, and the formula is: ; in, , The coordinates are in the muscle projection coordinate system. The electromyographic rate of change along the spine is denoted as α.

[0011] Preferably, in step S5, the formula for the dynamic adjacency matrix in the multimodal dynamic graph convolution model is: ; in, , Linear projection of channel features, The biomechanical prior matrix generated for spinal ligament tension. It is a row-normalized exponential function.

[0012] Preferably, in step S5, the formula for the spatiotemporal graph convolutional layer in the multimodal dynamic graph convolutional model is: ; in, For regularized Laplacian matrix, It is a Chebyshev polynomial basis. This is the ReLU activation function.

[0013] Preferably, in step S5, the multimodal dynamic graph convolutional model adopts the LSTM-Attention mechanism, and its attention weights are constrained by the Cobb angle rate of change, as shown in the formula: ; in, For attention weights, This is the spinal state vector. For LSTM hidden states, , This is a trainable weight matrix.

[0014] Preferably, in step S6, the muscle stiffness coefficient is estimated using the amplitude-frequency characteristics of surface electromyography signals, as shown in the formula: ; in, This represents the peak value of the electromyographic power spectrum. This represents the characteristic frequency threshold for muscle contraction. This is the frequency band energy integral.

[0015] The present invention achieves the following beneficial technical effects compared to the prior art: This invention provides a modeling method for the intrinsic correlation between paraspinal muscle function changes and scoliosis. The proposed phase-dependent bilateral muscle asymmetry index is the first to incorporate the spinal rotation phase into the assessment of electromyographic asymmetry, solving the problem that traditional static indicators cannot capture dynamic compensation. The multimodal dynamic graph convolutional model integrates learnable topology and biomechanical constraints, preserving the advantages of data-driven approaches while ensuring that the prediction results conform to the laws of spinal mechanics, and reducing the model's prediction error for the Cobb angle to an extremely low level. Finally, a real-time feedback system based on the muscle-spine differential equation can convert BAI (bilateral muscle asymmetry) into vibrational cue signals to guide patients' movement correction, which is significantly better than traditional training methods. This provides the first full-cycle solution for AIS (scoliosis-associated spinal musculoskeletal syndrome) that spans "pathological mechanism research - quantitative assessment - real-time intervention". Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 Flowchart of the modeling method for the intrinsic correlation between paravertebral muscle function changes and scoliosis provided by the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] The purpose of this invention is to provide a method for modeling the intrinsic correlation between changes in paraspinal muscle function and scoliosis, in order to solve the problems existing in the prior art.

[0020] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0021] Example 1: This embodiment provides a method for modeling the intrinsic correlation between changes in paravertebral muscle function and scoliosis, including the following steps: S1. Collect surface electromyography signals and synchronous spinal kinematic data of AIS patients while they are standing and bending over. The surface electromyography (EMG) signal includes 64 channels of data; the kinematic data includes the spinal rotation phase angle τ and Cobb angle. The kinematic data is acquired through an inertial measurement unit with a sampling frequency ≥100Hz and a time synchronization error with the surface EMG signal <5ms.

[0022] S2. Preprocess the surface electromyography signal by adaptive wavelet packet transform fusion with empirical mode decomposition, using the following formula: Used to eliminate motion artifacts and baseline drift in electromyography signals; Combined with EMD-WPT, non-stationary noise suppression is achieved. in, Let k be the kth order eigenmode function. This is an empirical mode decomposition operator used to decompose non-stationary signals into intrinsic mode functions. This represents the original surface electromyography signal. For the preprocessing results, For wavelet packet transform, These are Symlet-5 wavelet basis functions. For adaptive threshold, , Let N be the standard deviation of the k-th intrinsic mode function, N be the signal length, and K be the total number of effective IMF components. It also includes motion artifact removal, the formula is: This eliminates the distortion of electromyographic signals caused by bending over. in, The vertical motion coupling coefficient is... The horizontal damping factor is... Vertical acceleration, This represents the magnitude of the horizontal plane acceleration vector.

[0023] S3. Extract time-domain features, frequency-domain features, and spatial features from the preprocessed data; The time-domain feature is the improved weighted root mean square value, and the formula is: ; ; Enhance the signal contribution during key phases of the movement cycle (such as the maximal muscle contraction phase); in, Let i be the electromyographic amplitude at the i-th sampling point. The time window weighting function, The number of sampling points in the analysis window; The frequency domain characteristic is the electromyographic power spectrum entropy, and the formula is: The decrease in entropy reflects spectral compression caused by muscle fatigue. in, For the normalized power spectrum, The entropy of the electromyographic power spectrum; The spatial feature is the paraspinal muscle activation gradient, and the formula is: Spatial gradient used to quantify paraspinal muscle activation; in, , The coordinates are in the muscle projection coordinate system. The electromyographic rate of change along the spine is denoted as α.

[0024] S4. Calculate the phase-dependent bilateral muscle asymmetry index based on the three-dimensional features, using the following formula: ; in, The phase angle of spinal rotation. , These are the W-RMS values ​​for the left and right channels. This is the time decay factor, with a value of 0.15. To prevent the value from being zero, the value is set to 0.01.

[0025] S5. Construct a multimodal dynamic graph convolution model; The formula for the dynamic adjacency matrix in a multimodal dynamic graph convolution model is: Used to fuse data-driven connectivity and anatomical constraints; in, , Linear projection of channel features, The biomechanical prior matrix generated for spinal ligament tension. For row-normalized exponential functions; The formula for the spatiotemporal graph convolutional layer in a multimodal dynamic graph convolutional model is: Achieve spatiotemporal graph convolution to capture dynamic coupling of muscle networks; in, For regularized Laplacian matrix, It is a Chebyshev polynomial basis. It is the ReLU activation function; The multimodal dynamic graph convolutional model employs the LSTM-Attention mechanism, where the attention weights incorporate the Cobb angle rate of change constraint, as shown in the formula: Introducing a biomechanical prior-guided temporal attention mechanism; in, For attention weights, This is the spinal state vector. For LSTM hidden states, , This is a trainable weight matrix.

[0026] S6. Establish the joint differential equation for muscle-spine stiffness, the equation is as follows: Thus, the biomechanical differential equation of electromyography-spine was established for the first time. in, This is the muscle stiffness coefficient. Let be the ligament stiffness constant. For the angle of scoliosis, This represents the angle of the spine in a neutral position. It is the viscous damping coefficient; The muscle stiffness coefficient is estimated using the amplitude-frequency characteristics of surface electromyography (EMG) signals, and the formula is as follows: Achieve a positive correlation between muscle stiffness and spectral characteristics; in, This represents the peak value of the electromyographic power spectrum. This represents the characteristic frequency threshold for muscle contraction. This is the frequency band energy integral.

[0027] S7. Train the model using the physical constraint loss function, the formula is: Improve model interpretability through physical constraints; in, , These are the weighting coefficients. To predict the Cobb angle for the model, To measure the Cobb angle by X-ray, This is the right-hand side term of the differential equation.

[0028] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0029] It should be noted that the components mentioned in the above embodiments are all general standard parts or components known to those skilled in the art. Their structures and principles can be learned by those skilled in the art through technical manuals or conventional experimental methods.

[0030] This invention has illustrated its principles and implementation methods using specific examples. The descriptions of these embodiments are merely illustrative of the method and its core ideas; furthermore, those skilled in the art will recognize that modifications may be made to the specific implementation methods and application scope based on the principles of this invention. Therefore, the content of this specification should not be construed as limiting the invention.

Claims

1. A method for modeling the intrinsic correlation between changes in paravertebral muscle function and scoliosis, characterized in that: Includes the following steps: S1. Collect surface electromyography signals and synchronous spinal kinematic data of AIS patients while they are standing and bending over. S2. Preprocess the surface electromyography signal by adaptive wavelet packet transform fusion with empirical mode decomposition, using the following formula: ; ; in, Let k be the kth order eigenmode function. This is an empirical mode decomposition operator used to decompose non-stationary signals into intrinsic mode functions. This represents the original surface electromyography signal. For the preprocessing results, For wavelet packet transform, These are Symlet-5 wavelet basis functions. For adaptive threshold, , Let N be the standard deviation of the k-th intrinsic mode function, N be the signal length, and K be the total number of effective IMF components. S3. Extract time-domain features, frequency-domain features, and spatial features from the preprocessed data; S4. Calculate the phase-dependent bilateral muscle asymmetry index based on the three-dimensional features, using the following formula: ; in, The phase angle of spinal rotation. , These are the W-RMS values ​​for the left and right channels. This is the time decay factor, with a value of 0.

15. To prevent division by zero constant, a value of 0.01 is used; S5. Construct a multimodal dynamic graph convolution model; S6. Establish the joint differential equation for muscle-spine stiffness, the equation is as follows: ; in, This is the muscle stiffness coefficient. Let be the ligament stiffness constant. The angle of scoliosis. This represents the angle of the spine in a neutral position. It is the viscous damping coefficient; S7. Train the model using the physical constraint loss function, the formula is: ; in, , These are the weighting coefficients. To predict the Cobb angle for the model, To measure the Cobb angle by X-ray, This is the right-hand side term of the differential equation.

2. The method for modeling the intrinsic correlation between paravertebral muscle function changes and scoliosis according to claim 1, characterized in that: In step S1, the surface electromyography signal includes 64 channels of data.

3. The method for modeling the intrinsic correlation between paravertebral muscle function changes and scoliosis according to claim 1, characterized in that: In step S1, the kinematic data includes the spinal rotation phase angle τ and the Cobb angle, and the kinematic data is acquired through an inertial measurement unit.

4. The method for modeling the intrinsic correlation between paravertebral muscle function changes and scoliosis according to claim 1, characterized in that: Step S2 also includes motion artifact removal, with the following formula: ; in, The vertical motion coupling coefficient is... The horizontal damping factor is... Vertical acceleration, This represents the magnitude of the horizontal plane acceleration vector.

5. The method for modeling the intrinsic correlation between paravertebral muscle function changes and scoliosis according to claim 1, characterized in that: In step S3, the time-domain feature is the improved weighted root mean square value, and the formula is: ; ; in, Let i be the electromyographic amplitude at the i-th sampling point. The time window weighting function, The number of sampling points in the analysis window; The frequency domain feature is the electromyographic power spectrum entropy, and the formula is: ; in, For the normalized power spectrum, The entropy of the electromyographic power spectrum; The spatial feature is the paraspinal muscle activation gradient, and the formula is: ; in, , The coordinates are in the muscle projection coordinate system. The electromyographic rate of change along the spine is denoted as α.

6. The method for modeling the intrinsic correlation between paravertebral muscle function changes and scoliosis according to claim 1, characterized in that: In step S5, the formula for the dynamic adjacency matrix in the multimodal dynamic graph convolution model is: ; in, , Linear projection of channel features, The biomechanical prior matrix generated for spinal ligament tension. It is a row-normalized exponential function.

7. The method for modeling the intrinsic correlation between paravertebral muscle function changes and scoliosis according to claim 1, characterized in that: In step S5, the formula for the spatiotemporal graph convolutional layer in the multimodal dynamic graph convolutional model is: ; in, For regularized Laplacian matrix, It is a Chebyshev polynomial basis. This is the ReLU activation function.

8. The method for modeling the intrinsic correlation between paravertebral muscle function changes and scoliosis according to claim 1, characterized in that: In step S5, the multimodal dynamic graph convolutional model adopts the LSTM-Attention mechanism, and its attention weights are constrained by the Cobb angle rate of change, as shown in the formula: ; in, For attention weights, This is the spinal state vector. For LSTM hidden states, , This is a trainable weight matrix.

9. The method for modeling the intrinsic correlation between paravertebral muscle function changes and scoliosis according to claim 1, characterized in that: In step S6, the muscle stiffness coefficient is estimated using the amplitude-frequency characteristics of surface electromyography signals, as shown in the formula: ; in, This represents the peak value of the electromyographic power spectrum. This represents the characteristic frequency threshold for muscle contraction. This is the frequency band energy integral.

Citation Information

Patent Citations

  • Scoliosis assessment method and system based on multi-modal data

    CN120093221A