A method of detecting muscle function for assessing gastrocnemius muscle recoil force characteristics

Abstract

The present application relates to the technical field of muscle biomechanics evaluation, in particular to a detection method for evaluating gastrocnemius muscle rebound force characteristics, comprising: separating out active and passive components and fusing into a composite feature vector by decomposing collected physiological and physical signals. According to the gastrocnemius muscle anatomical structure, the body surface projection area is virtually divided into feature analysis blocks with hierarchical structure and weighted, and the overall dynamic response index is calculated accordingly. Combined with the gastrocnemius biomechanics simulation model, the instant tension distribution of internal muscle fiber bundles at different depths is reversely solved, and then the elastic potential energy storage and release efficiency is calculated, and the rebound force characteristic grade is determined by comparing with the preset threshold. The method realizes the accurate evaluation based on the partition of anatomical structure, and obtains the deep mechanical state of muscle fiber bundle level through model reverse solving, which improves the anatomical accuracy and mechanical depth of evaluation.

Description

Technical Field

[0001] This invention relates to the field of muscle biomechanical assessment technology, and in particular to a method for detecting muscle function to assess the rebound force characteristics of the gastrocnemius muscle. Background Technology

[0002] Existing gastrocnemius muscle function assessment techniques largely rely on surface electromyography, isokinetic muscle strength testing, or ultrasound imaging. These methods can acquire overall electrical signals or macroscopic mechanical parameters of muscle activity, but they struggle to separate active contraction from passive kinetic characteristics. Conventional signal processing techniques, such as filtering or time-frequency analysis, can extract some features but cannot effectively distinguish the independent contributions of active and passive muscle components. In terms of regional analysis, existing methods often use uniform grids or empirical surface partitioning, failing to consider the specificity of the gastrocnemius muscle's anatomy, resulting in a lack of physiological basis for local functional assessment.

[0003] Current technologies cannot quantify the instantaneous tension state of muscle fiber bundles at different depths within the gastrocnemius muscle. Traditional biomechanical models are mostly used for forward simulation of muscle contraction, but they struggle to inversely deduce the mechanical distribution of deep muscles from complex signals collected from the body surface. This stems from the fact that model construction typically ignores the hierarchical nature of muscle fiber orientation and lacks effective algorithms to correlate surface signals with internal tension. Therefore, existing methods cannot accurately assess the efficiency of elastic potential energy storage and release in the gastrocnemius muscle during movement, limiting the reliability and clinical application value of rebound force characteristic analysis.

[0004] This invention aims to address how to achieve precise regional weighting based on the anatomical structure of the gastrocnemius muscle to more accurately reflect the functional contributions of different regions. Simultaneously, it seeks to overcome the technical challenge of inversely solving the tension distribution at the muscle fiber bundle level from surface signals, in order to achieve non-invasive assessment of deep muscle biomechanics. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and to propose a method for detecting muscle function to evaluate the rebound force characteristics of the gastrocnemius muscle.

[0006] To achieve the above objectives, the present invention employs the following technical solution: a method for detecting muscle function to evaluate the rebound force characteristics of the gastrocnemius muscle, comprising: The raw physiological and physical signals generated by the gastrocnemius muscle of the subject under specific exercise conditions were collected and the signals were decomposed to separate the active component characterizing the active contraction characteristics of the gastrocnemius muscle and the passive component characterizing the passive dynamic characteristics of the gastrocnemius muscle. The separated active and passive components are temporally aligned and feature-fused to construct a composite feature vector. Based on the anatomical structure and muscle fiber orientation of the gastrocnemius muscle, the corresponding body surface area is virtually divided into multiple hierarchical feature analysis blocks, and a weight coefficient is assigned to each feature analysis block. The state of each feature analysis block is assigned using the fused composite feature vector, and the overall dynamic response index of the gastrocnemius muscle is calculated based on the weight coefficient of each feature analysis block. Based on the overall dynamic response index, and combined with the pre-established gastrocnemius muscle biomechanical simulation model, the instantaneous tension distribution of muscle fiber bundles at different depths inside the gastrocnemius muscle under the specific motion state is solved in reverse. Based on the instantaneous tension distribution, the inflection point and periodic peak of tension change are identified, and then the elastic potential energy storage and release efficiency of the gastrocnemius muscle in the continuous contraction and relaxation cycle is calculated. The elastic potential energy storage and release efficiency is compared with multiple preset gastrocnemius muscle function level thresholds to determine the rebound force characteristic level of the subject's gastrocnemius muscle, thus completing the detection and evaluation of gastrocnemius muscle function.

[0007] Preferably, the step of performing signal decomposition to separate the active component characterizing the active contraction characteristics of the gastrocnemius muscle and the passive component characterizing the passive dynamic characteristics of the gastrocnemius muscle specifically includes: The original physiological signals include at least electromyographic signals, and the original physical signals include at least pressure signals and position signals; After performing multi-scale wavelet transform on the electromyographic signal to filter out electrocardiographic interference and motion artifacts, the neural activation sequence representing the recruitment and firing frequency of motor units was extracted as the active component. The pressure signal and the position signal are jointly analyzed. Baseline drift is removed by high-pass filtering. Then, the first and second derivatives of the signals are calculated to extract physical parameters that reflect the deformation rate, deformation acceleration and ground reaction force of the gastrocnemius muscle. The physical parameters are combined into passive components. A time delay correction relationship is established between the active and passive components. The conduction time from neural activation to mechanical response is determined through cross-correlation analysis. Based on this, the active component is time-adjusted in advance to complete the time-domain alignment.

[0008] Preferably, constructing the composite feature vector further includes: The time-aligned neural activation sequence is spliced ​​with the physical parameter sequence to form the initial fusion sequence; Principal component analysis is performed on the initial fusion sequence, and the first few principal components whose contribution rates exceed a preset threshold are retained to form a dimensionality-reduced feature subset. The principal components in the feature subset are respectively compared with the baseline features extracted from the gastrocnemius muscle signal of the subject at rest to obtain the dynamic feature increment after removing individual static differences. The dynamic feature increments are normalized, and the final output is the composite feature vector. The composite feature vector can simultaneously reflect the nerve activation state and mechanical response of the gastrocnemius muscle.

[0009] Preferably, the step of virtually dividing the body surface region corresponding to the gastrocnemius muscle into multiple feature analysis blocks with a hierarchical structure, and assigning a weight coefficient to each feature analysis block, includes: Obtain ultrasound or magnetic resonance images of the gastrocnemius muscle region of the subject, and mark the boundary between the medial and lateral heads of the gastrocnemius muscle, the location of the muscle belly, and the Achilles tendon junction area on the images; Based on the annotation information, the entire gastrocnemius muscle projection area is divided into three levels: the core area, the transition area, and the edge area. Within the core area, it is further subdivided into longitudinal muscle fiber-dominated areas and oblique muscle fiber-dominated areas based on the direction of the muscle bundles. Each segmented feature analysis block is assigned a weight coefficient, with the core area having a higher weight coefficient than the transition area, the transition area having a higher weight coefficient than the edge area, and the longitudinal muscle fiber dominant area having a higher weight coefficient than the oblique muscle fiber dominant area.

[0010] Preferably, the step of calculating the overall dynamic response index of the gastrocnemius muscle based on the weight coefficient of each feature analysis block includes: Each feature element in the composite feature vector is mapped and assigned according to its physiological meaning and the functional characteristics of the feature analysis block to obtain the feature value of each feature analysis block. The weighted contribution value of each feature analysis block is obtained by multiplying the feature value of each feature analysis block by the weight coefficient of the feature analysis block. The weighted contribution values ​​of all feature analysis blocks are superimposed linearly or nonlinearly, and the formula or model parameters of the operation are obtained by training and calibration on test data of the standard population in the early stage. The result of the superposition operation is scaled and mapped to the range of zero to one hundred, and the output is the overall dynamic response index.

[0011] Preferably, the inverse solution of the instantaneous tension distribution of muscle fiber bundles at different depths within the gastrocnemius muscle under the specific motion state includes: The gastrocnemius muscle biomechanical simulation model is a multi-scale finite element model that includes the epimysium, perimysium, muscle fibers, series elastic elements, and parallel elastic elements. The overall dynamic response index is used as the overall driving input of the gastrocnemius muscle biomechanical simulation model. In the aforementioned gastrocnemius muscle biomechanical simulation model, differentiated material property parameters and initial preload are set for muscle fiber bundles at different depths; The gastrocnemius muscle biomechanical simulation model is run for forward calculation. The surface mechanical prediction value output by the model is compared with the actual collected pressure signal. The activation level parameters of the muscle fiber bundles inside the model are adjusted by iterative optimization algorithm until the error between the predicted value and the actual value is less than the allowable range. The stress field data of each muscle fiber bundle within the optimized model are extracted and spatially integrated to obtain the instantaneous tension distribution of each muscle fiber bundle.

[0012] Preferably, the step of running the gastrocnemius muscle biomechanical simulation model for forward calculation involves comparing the surface mechanical prediction values ​​output by the model with the actual collected pressure signals, and adjusting the activation level parameters of the muscle fiber bundles within the model through an iterative optimization algorithm until the error between the predicted and actual values ​​is less than the allowable range, including: The overall dynamic response index is input into the gastrocnemius muscle biomechanical simulation model, and the forward calculation process of the model is executed to obtain the theoretical pressure value of each measurement point on the surface of the gastrocnemius muscle, which is used as the surface mechanical prediction value. Simultaneously acquire the pressure signal values ​​of each corresponding measurement point that are actually collected; Calculate the absolute difference between the predicted surface mechanical value and the actual collected pressure signal value at each measurement point; The total prediction error is obtained by summing the weighted absolute differences of all measurement points. Determine whether the overall prediction error is less than a preset allowable error range; When the overall prediction error is greater than or equal to the preset allowable error range, the gradient descent algorithm is used to adjust the activation level parameters of each muscle fiber bundle inside the gastrocnemius muscle biomechanical simulation model. Repeat the forward calculation, error calculation, and parameter adjustment steps until the overall prediction error is less than the preset allowable error range.

[0013] Preferably, the step of extracting and spatially integrating the stress field data of each muscle fiber bundle within the optimized model to obtain the instantaneous tension distribution of each muscle fiber bundle includes: When the overall prediction error is less than the preset allowable error range, the activation level parameters of each muscle fiber bundle inside the current gastrocnemius muscle biomechanical simulation model are locked. The stress tensor data of each muscle fiber bundle in three-dimensional space are extracted from the locked gastrocnemius muscle biomechanical simulation model to form stress field data; For each muscle fiber bundle, the stress tensor data is calculated as a one-dimensional line integral along the direction of the muscle fiber bundle within its spatial domain. The integral calculation result is used as the instantaneous tension value of the muscle fiber bundle in the current motion state; The instantaneous tension values ​​of all muscle fiber bundles are collected to form the instantaneous tension distribution, which reflects the distribution of tension magnitude within the gastrocnemius muscle.

[0014] Preferably, the step of further calculating the elastic potential energy storage and release efficiency of the gastrocnemius muscle in a continuous contraction-relaxation cycle includes: For the curve of the instantaneous tension distribution of each muscle fiber bundle over time, identify the tension rise and tension fall segments within each movement cycle. Calculate the area under the tension rise curve, where the area represents the elastic potential energy stored in the muscle fiber bundle during each movement cycle; Calculate the area under the tension decrease curve, where the area represents the elastic potential energy released by the muscle fiber bundle in each movement cycle; The ratio of released elastic potential energy to stored elastic potential energy is defined as the single-cycle efficiency of the muscle fiber bundle. The overall elastic potential energy storage and release efficiency of the gastrocnemius muscle is obtained by weighting the individual cycle efficiency of all muscle fiber bundles according to their cross-sectional area.

[0015] Preferably, determining the rebound force characteristic level of the subject's gastrocnemius muscle includes: A threshold database of gastrocnemius muscle function levels was established by testing a large number of healthy people and people with different degrees of muscle dysfunction in advance. The database divides the numerical range of elastic potential energy storage and release efficiency into five levels: excellent, good, moderate, damaged, and severely damaged. The calculated values ​​of the elastic potential energy storage and release efficiency of the subject's gastrocnemius muscle as a whole are matched with the range in the gastrocnemius muscle functional level threshold database. Output the matched functional level labels, and generate a structured evaluation report that includes the contribution of each feature analysis block and the main limiting factors.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: Based on the anatomical structure and muscle fiber orientation of the gastrocnemius muscle, the body surface area is virtually divided into multiple hierarchical feature analysis blocks, and a weight coefficient is assigned to each block. This technique takes into account the physiological structure of the gastrocnemius muscle, making the partitioning more closely resemble the actual distribution of muscle function. The introduction of weight coefficients quantifies the contribution of each block to overall function based on the physiological roles of different blocks in contraction and passive stretching. Through this partitioning and weighting method, when calculating the overall dynamic response index from the composite feature vector, the true state of the muscle can be more accurately represented, reducing the evaluation bias caused by uniform partitioning. This improves the anatomical accuracy of regional functional analysis, allowing the subsequent overall index to better reflect the complex dynamic behavior of the gastrocnemius muscle.

[0017] Using a pre-established biomechanical simulation model of the gastrocnemius muscle, the instantaneous tension distribution of muscle fiber bundles at different depths is solved inversely based on the overall dynamic response index. This technique utilizes the muscle geometry, material properties, and contraction kinetic parameters included in the simulation model to establish a mathematical correlation between surface signals and internal mechanical states. Through the inverse solution algorithm, the macroscopic index is transformed into tension data of microscopic muscle fiber bundles. This process achieves non-invasive acquisition of real-time mechanical information of deep muscles, overcoming the limitations of conventional forward simulation. This provides a tension distribution picture at the muscle fiber bundle level, enabling the calculation of elastic potential energy storage and release efficiency based on a more refined mechanical foundation. This enhances the depth and reliability of the rebound force characteristic assessment, providing a more direct mechanical basis for functional level determination. Attached Figure Description

[0018] Figure 1 This is a flowchart of the muscle function testing method for evaluating the rebound force characteristics of the gastrocnemius muscle according to the present invention; Figure 2 A flowchart for constructing composite feature vectors; Figure 3 A flowchart for feature analysis block division and weighting; Figure 4 A three-stage time series characteristic diagram of the overall dynamic response index of the gastrocnemius muscle; Figure 5 The curves show the tension changes and elastic potential energy storage-release of the gastrocnemius muscle during a single movement cycle. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0020] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the 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, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0021] Please see Figure 1 This invention proposes a method for detecting muscle function, specifically assessing the rebound force characteristics of the gastrocnemius muscle. The overall implementation scheme is as follows: During the detection process, raw physiological and physical signals generated by the gastrocnemius muscle of the subject under specific movement states are collected, such as continuous jumping or tiptoeing at a predetermined speed. The collected mixed signals are decomposed to separate the active component characterizing the active contraction characteristics of the gastrocnemius muscle and the passive component characterizing its passive dynamic characteristics. The two separated components are temporally aligned and feature fusion is performed to construct a composite feature vector. Based on the anatomical structure of the gastrocnemius muscle, its corresponding surface projection area is virtually divided into multiple hierarchical feature analysis blocks with different weights. The composite feature vector is used to assign state values ​​to each block, and combined with its weight coefficients, an overall dynamic response index characterizing the overall response intensity of the gastrocnemius muscle is calculated. This index is used as input to drive a pre-established gastrocnemius muscle biomechanical simulation model, and the model is used to inversely solve for the instantaneous tension distribution of muscle fiber bundles at different depths within the gastrocnemius muscle under movement. By analyzing the curve of tension distribution over time, inflection points and peak values ​​are identified, and the efficiency of elastic potential energy storage and release in the gastrocnemius muscle during continuous contraction and relaxation cycles is calculated. This efficiency value is compared with a preset functional level threshold to determine the rebound force characteristic level of the subject's gastrocnemius muscle, thereby completing the quantitative detection and evaluation of muscle function.

[0022] In one embodiment of the present invention, see [reference] Figure 2The collected raw physiological signals include at least electromyography (EMG) signals, and the raw physical signals include at least pressure and position signals. Multi-scale wavelet transform is performed on the EMG signals to filter out ECG interference and motion artifacts, and neural activation sequences representing the recruitment and firing frequencies of motor units are extracted as active components. Joint analysis of pressure and position signals is performed, using high-pass filtering to remove baseline drift. Then, by calculating the first and second derivatives of the signals, physical parameters reflecting the deformation rate, deformation acceleration, and changes in ground reaction force of the gastrocnemius muscle are extracted and combined into passive components. A time delay correction relationship is established between the active and passive components. Cross-correlation analysis is used to determine the transmission time from neural activation to mechanical response, and time advance compensation is applied to the neural activation sequences accordingly to achieve time-domain alignment. The time-aligned neural activation sequences are concatenated with the physical parameter sequences to form an initial fusion sequence. Principal component analysis is performed on this initial fusion sequence, retaining the top few principal components with contribution rates exceeding a preset threshold to form a dimensionality-reduced feature subset. Each principal component in this feature subset is differentially calculated with the baseline features extracted from the gastrocnemius muscle signal at the subject's resting state to obtain dynamic feature increments that remove individual static differences. These dynamic feature increments are then normalized, and the final output is a composite feature vector that synchronously reflects the gastrocnemius muscle's neural activation state and mechanical response.

[0023] In specific implementation, the embodiments involve the acquisition, decomposition, temporal alignment, and feature fusion of raw physiological and physical signals. Electromyographic (EMG) signals, pressure signals, and position signals constitute the acquired raw signal set. Multi-scale wavelet transform is performed on the EMG signals to filter out ECG interference and motion artifacts. The extracted neural activation sequences characterize the recruitment and firing frequencies of motor units and are used as active components. Joint analysis is performed on the pressure and position signals. After removing baseline drift through high-pass filtering, the first and second derivatives of the signals are calculated. The extracted physical parameters include the gastrocnemius muscle deformation rate, deformation acceleration, and changes in ground reaction force. These physical parameters are combined to form passive components. The time delay correction relationship between active and passive components is established through cross-correlation analysis. Cross-correlation analysis determines the transmission time from neural activation to mechanical response and performs time advance compensation on the neural activation sequence accordingly, thus completing temporal alignment.

[0024] In some embodiments, the time-domain aligned neural activation sequence and the physical parameter sequence are spliced ​​together to form an initial fusion sequence. Principal component analysis is performed on the initial fusion sequence, and the first few principal components with contribution rates exceeding a preset threshold are retained to form a dimensionality-reduced feature subset. The principal components in the feature subset are then differentially calculated with the baseline features extracted from the gastrocnemius muscle signal of the subject at rest to obtain dynamic feature increments that remove individual static differences. The dynamic feature increments are then normalized to finally output a composite feature vector, which synchronously reflects the gastrocnemius muscle nerve activation state and mechanical response.

[0025] Optionally, the normalization process uses the minimum-maximum scaling method, expressed by the formula: in: Represents any feature value in the dynamic feature increment. This represents the minimum value of the feature in the training dataset. This represents the maximum value of the feature in the training dataset. This represents the normalized feature value. The training dataset is derived from previous test data on a standard population.

[0026] Understandably, the multi-scale wavelet transform uses the Daubechies wavelet basis function, with a preset threshold of 85% of the cumulative contribution rate of the principal components. Electromyography (EMG) signals are acquired using surface electrodes attached to the belly of the gastrocnemius muscle. Pressure signals are acquired using a pressure sensor array placed on the sole of the foot. Position signals are acquired using an optical motion capture system that tracks markers on the lower leg and foot. ECG interference filtering is based on a threshold denoising method using wavelet transform. Motion artifact filtering is based on an adaptive filtering algorithm. When jointly analyzing pressure and position signals, the first derivative is calculated to obtain the deformation rate, and the second derivative is calculated to obtain the deformation acceleration. Changes in ground reaction force are obtained by integrating the pressure signal and calibrating it synchronously with the position signal.

[0027] In some embodiments, cross-correlation analysis calculates the maximum cross-correlation coefficient between the neural activation sequence and the physical parameter sequence to determine the conduction time. Time advance compensation shifts the neural activation sequence forward along the time axis by the number of sample points corresponding to the conduction time. The initial fusion sequence is spliced ​​by concatenating the neural activation sequence and the physical parameter sequence into a long vector according to the time points. Before principal component analysis, the initial fusion sequence is zero-mean processed. Baseline features are extracted from electromyographic signals, pressure signals, and position signals collected when the subject is standing still. Difference calculation subtracts the principal component value of the corresponding baseline feature from each principal component of the feature subset. The dynamic feature increment reflects the change in the motion state relative to the resting state.

[0028] It is understandable that the dimension of the composite feature vector is equal to the number of principal components retained. The composite feature vector is used for the state assignment of the subsequent feature analysis blocks. The extraction of neural activation sequences is based on the detection of the signal envelope after wavelet coefficient reconstruction. The extraction of physical parameter sequences is based on the derivative operation of the filtered signal. Temporal alignment ensures that the active and passive components are synchronized on the time axis. The feature fusion process integrates neural and mechanical information to comprehensively characterize the gastrocnemius muscle function.

[0029] In one embodiment of the present invention, see [reference] Figure 3 Ultrasound or magnetic resonance imaging (MRI) images of the gastrocnemius muscle region of the subjects were acquired, and the boundaries between the medial and lateral heads of the gastrocnemius muscle, the location of the muscle belly, and the Achilles tendon junction were marked on the images. Based on the marked information, the entire gastrocnemius muscle projection area was divided into three levels: a core area, a transition area, and a peripheral area. Within the core area, it was further subdivided into longitudinal muscle fiber-dominated areas and oblique muscle fiber-dominated areas according to the direction of the muscle fibers. Weight coefficients were assigned to each segmented feature analysis block, with the weight coefficient of the core area being higher than that of the transition area, the weight coefficient of the transition area being higher than that of the peripheral area, and the weight coefficient of the longitudinal muscle fiber-dominated area being higher than that of the oblique muscle fiber-dominated area. Each feature element in the composite feature vector was mapped and assigned according to its physiological meaning and the functional characteristics of the feature analysis block to obtain the feature value of each feature analysis block. The feature value of each feature analysis block was multiplied by its weight coefficient to obtain the weighted contribution value of that block. A linear or nonlinear superposition operation was performed on the weighted contribution values ​​of all feature analysis blocks. The formula or model parameters for the operation were obtained by training and calibration using test data from a standard population in the early stage. The result of the superposition operation is scaled and mapped to the range of 0 to 100, and the output is the overall dynamic response index.

[0030] In practice, the process involves dividing the gastrocnemius muscle into characteristic analysis blocks based on its anatomical structure and calculating the overall dynamic response index. Ultrasound or magnetic resonance images of the gastrocnemius muscle region of the subject are obtained. The boundaries between the medial and lateral heads of the gastrocnemius muscle, the position of the muscle belly, and the Achilles tendon junction area are marked on the ultrasound or magnetic resonance images. Based on the marked information, the entire gastrocnemius muscle projection area is divided into three levels: the core area, the transition area, and the peripheral area. Within the core area, it is further subdivided into longitudinal muscle fiber dominant area and oblique muscle fiber dominant area according to the direction of the muscle fibers. A weight coefficient is assigned to each divided characteristic analysis block. The weight coefficient of the core area is higher than that of the transition area, the weight coefficient of the transition area is higher than that of the peripheral area, and the weight coefficient of the longitudinal muscle fiber dominant area is higher than that of the oblique muscle fiber dominant area.

[0031] In some embodiments, each feature element in the composite feature vector is mapped and assigned according to its physiological meaning and the functional characteristics of the feature analysis block. The mapping and assignment are completed based on a predefined lookup table. The lookup table defines the association between neural activation features of different frequencies and contraction characteristics of specific muscle fiber regions, as well as the association between different physical parameters and mechanical deformation of specific regions. The feature value of each feature analysis block is obtained through mapping and assignment. The feature value of each feature analysis block is multiplied by the weight coefficient of the feature analysis block to obtain the weighted contribution value of the feature analysis block.

[0032] Optionally, a nonlinear superposition operation is performed on the weighted contribution values ​​of all feature analysis blocks to calculate the overall dynamic response index. The nonlinear superposition operation adopts a fusion model based on radial basis functions, and the formula is expressed as follows: in: Represents the overall dynamic response index. The total number of blocks representing feature analysis. Representing the The weight coefficients of each feature analysis block. Representing the Feature values ​​of each feature analysis block Representing the Reference center value of feature values ​​of each feature analysis block. The model parameters representing shape parameters and nonlinear superposition operations are obtained through prior training and calibration on test data of a standard population. During the training process, the least squares method is used to fit the relationship between the overall dynamic response index of the standard population and the expert scores.

[0033] It is understood that the acquisition of ultrasound or MRI images adopts a standardized limb positioning and scanning protocol. The core area corresponds to the thickest part of the gastrocnemius muscle belly and is the area with the clearest muscle fiber structure on ultrasound or MRI images. The transition area surrounds the core area and connects to the Achilles tendon junction area. The peripheral area covers the remaining peripheral part of the gastrocnemius muscle projection area. The determination of muscle fiber direction is based on the imaging direction of muscle fiber stripes on ultrasound or MRI images. The longitudinal muscle fiber dominant area refers to the area where the angle between the muscle fiber direction and the long axis of the muscle is less than 15 degrees. The oblique muscle fiber dominant area refers to the area where the angle between the muscle fiber direction and the long axis of the muscle is between 15 and 45 degrees. The specific values ​​of the weight coefficients are preset based on quantitative studies of the contribution of different regions to the strength of the gastrocnemius muscle in anatomical literature. The preset range of the weight coefficient for the core area is between 0.4 and 0.6, the preset range of the weight coefficient for the transition area is between 0.2 and 0.4, and the preset range of the weight coefficient for the peripheral area is between 0.05 and 0.15.

[0034] In some embodiments, the scaling transformation of the overall dynamic response index maps the result of the nonlinear superposition operation to the range of zero to one hundred. The scaling transformation adopts a linear scaling method. The scaling coefficient and offset are determined by statistically analyzing the maximum and minimum values ​​of the operation results in the standard population test data. The output overall dynamic response index is used to drive the gastrocnemius muscle biomechanical simulation model. The division of feature analysis blocks, weight allocation, and feature value mapping are all automatically completed in the dedicated image processing and calculation module. The image processing and calculation module integrates image segmentation algorithm and feature matching algorithm.

[0035] It is understandable that the feature analysis block is a virtually defined geometric region rather than a physical marker. The boundary information and weight coefficient information of the feature analysis block are stored in a configuration file or database. The calculation process of the overall dynamic response index is executed in real time during each subject test. The preliminary test data for the standard population includes signal data and corresponding clinical function scores of healthy individuals at different exercise intensities. The training and calibration process determines the parameters in the nonlinear superposition calculation formula. and The specific numerical value is that the exponential term in the radial basis function is used to measure the degree to which the feature value of each feature analysis block deviates from its ideal reference center value.

[0036] In one embodiment of the present invention, the gastrocnemius muscle biomechanical simulation model used is a multi-scale finite element model comprising the epimysium, perimysium, muscle fibers, series elastic elements, and parallel elastic elements. The calculated overall dynamic response index is used as the overall driving input of the biomechanical simulation model. Within the model, differentiated material property parameters and initial preload are set for muscle fiber bundles at different depths. The model is run for forward calculation, and the surface mechanical prediction values ​​output by the model are compared with the actual collected pressure signals. The activation level parameters of the muscle fiber bundles within the model are adjusted through an iterative optimization algorithm until the error between the predicted and actual values ​​is less than the allowable range. The stress field data of each muscle fiber bundle within the optimized model are extracted and spatially integrated to obtain the instantaneous tension distribution of each muscle fiber bundle.

[0037] In practice, the gastrocnemius muscle biomechanical simulation model is used to reverse engineer the instantaneous tension distribution of muscle fiber bundles. The gastrocnemius muscle biomechanical simulation model used is a multi-scale finite element model that includes the epimysium, perimysium, muscle fibers, series elastic elements, and parallel elastic elements. This model is constructed in commercial finite element software or a custom computational mechanics program. The geometric contour is established based on standard adult gastrocnemius muscle anatomy atlases or subject-specific medical imaging data. The epimysium and perimysium are modeled as hyperelastic soft tissue materials, and the muscle fibers are modeled as rod elements capable of active contraction. The series elastic elements and parallel elastic elements are integrated in the muscle fiber element model in the form of nonlinear springs and dampers to characterize the viscoelastic properties of the muscle.

[0038] In some embodiments, the calculated overall dynamic response index is used as the overall driving input of the gastrocnemius muscle biomechanical simulation model. The overall dynamic response index is converted into the load coefficient or the basic scalar of the preset activation level in the model boundary conditions. Different material property parameters and initial preload are set for muscle fiber bundles of different depths in the gastrocnemius muscle biomechanical simulation model. The preset value of the material elastic modulus of the superficial muscle fiber bundle is lower than that of the deep muscle fiber bundle, and the preset value of the initial preload of the deep muscle fiber bundle is higher than that of the superficial muscle fiber bundle. The specific values ​​of the material property parameters and the initial preload are derived from in vitro experimental data or previous model calibration results in muscle biomechanics literature.

[0039] Optionally, a gastrocnemius muscle biomechanical simulation model is run for forward calculation to solve for the predicted surface mechanical values. The forward calculation is performed based on a static or quasi-static solver of the finite element method. The predicted surface mechanical values ​​output by the model are compared with the actual collected pressure signals. The activation level parameters of the muscle fiber bundles inside the model are adjusted through an iterative optimization algorithm until the error between the predicted and actual values ​​is less than the allowable range. The activation level parameters of the muscle fiber bundles are coefficients that scale their maximum active contractile stress. Adjusting the activation level parameters directly changes the contractile force contribution of each muscle fiber bundle in the finite element calculation. The allowable error range is set to five percent of the average amplitude of the actual collected pressure signal.

[0040] It is understandable that extracting the instantaneous tension distribution from the optimized model involves post-processing calculations. This involves extracting the stress field data of each muscle fiber bundle within the gastrocnemius muscle biomechanical simulation model. This stress field data includes the Cauchy stress tensor at the integration point of each muscle fiber unit in the model. For the stress tensor data of each muscle fiber bundle, a one-dimensional line integral is calculated along the direction of the muscle fiber bundle within its spatial domain. The integral formula is expressed as: in: Representing the Instantaneous tension value of a strip muscle fiber bundle. Representing the The central axis path of the muscle fiber bundle Represents along the path Upper position Normal stress in the direction of muscle fibers at that location. Represents along the path Upper position The cross-sectional area of ​​the muscle fiber bundles at a given location is used to collect the instantaneous tension values ​​of all muscle fiber bundles, forming an instantaneous tension distribution that reflects the distribution of tension magnitude within the gastrocnemius muscle.

[0041] In some embodiments, the iterative optimization algorithm employs the Levenberg-Marquardt nonlinear least squares algorithm. This algorithm minimizes the sum of squared residuals between the predicted surface mechanics values ​​and the actual collected pressure signals by iteratively updating the activation level parameter vector of the myofibril bundles. Each iteration requires rerunning a complete forward calculation of the gastrocnemius muscle biomechanical simulation model to update the predicted surface mechanics values. Optimization stops when the sum of squared residuals is less than a preset threshold or the parameter change is less than a set tolerance. The optimization process is automatically completed in a dedicated numerical optimization calculation module.

[0042] It is understandable that muscle fiber bundles are grouped and defined as different sets of elements in the model. Each set of elements represents a muscle fiber bundle with the same depth properties and preset material parameters. The extraction of stress field data is achieved by accessing the result file of the finite element software or directly reading the memory data of the custom program. One-dimensional line integral calculation is performed on the pre-defined muscle fiber bundle path. The path is defined by a series of continuous node coordinates. The instantaneous tension distribution is output in the form of a vector or list, which contains the number of each muscle fiber bundle and its corresponding instantaneous tension value. The instantaneous tension distribution is used for subsequent calculation of elastic potential energy storage and release efficiency.

[0043] In one embodiment of the present invention, the overall dynamic response index is input into the gastrocnemius muscle biomechanical simulation model, and the forward calculation process of the model is executed to obtain the theoretical pressure values ​​of each measurement point on the surface of the gastrocnemius muscle, which are used as the surface mechanical prediction values. Simultaneously, the pressure signal values ​​of each corresponding measurement point are acquired. The absolute difference between the surface mechanical prediction value and the actual acquired pressure signal value at each measurement point is calculated. The absolute differences of all measurement points are weighted and summed to obtain the overall prediction error. It is determined whether the overall prediction error is less than a preset allowable error range. When the overall prediction error is greater than or equal to the preset allowable error range, the activation level parameters of each muscle fiber bundle within the model are adjusted using a gradient descent algorithm. The forward calculation, error calculation, and parameter adjustment steps are repeated until the overall prediction error is less than the preset allowable error range. When the overall prediction error is less than the preset allowable error range, the activation level parameters of each muscle fiber bundle within the current model are locked. The stress tensor data of each muscle fiber bundle in three-dimensional space is extracted from the locked model to form stress field data. For each muscle fiber bundle, a one-dimensional line integral is performed on the stress tensor data along the direction of the muscle fiber bundle within its spatial domain. The integral calculation result is used as the instantaneous tension value of the muscle fiber bundle under the current motion state. The instantaneous tension values ​​of all muscle fiber bundles are collected to form an instantaneous tension distribution that reflects the distribution of tension magnitude within the gastrocnemius muscle.

[0044] In practice, the process involves forward calculation, error comparison, parameter adjustment, and instantaneous tension distribution calculation of the gastrocnemius muscle biomechanical simulation model. The overall dynamic response index is input into the gastrocnemius muscle biomechanical simulation model, and the forward calculation process of the model is executed to obtain the theoretical pressure values ​​of each measurement point on the surface of the gastrocnemius muscle as the surface mechanical prediction values. The forward calculation process is completed based on the finite element solver. The solver calculates the stress and strain field of the model under quasi-static equilibrium according to the input boundary conditions and material properties and outputs the surface contact pressure. Simultaneously, the pressure signal values ​​of each measurement point are actually collected. The pressure signal values ​​are actually collected from the synchronously recorded data of the sensor array arranged on the sole of the foot or in contact with the gastrocnemius muscle under specific motion states.

[0045] In some embodiments, the absolute difference between the predicted surface mechanical value and the actual collected pressure signal value at each measurement point is calculated. The absolute differences at all measurement points are then weighted and summed to obtain the overall prediction error. The weight of each measurement point is allocated according to the weight coefficient of its corresponding feature analysis block. Measurement points in the core area have a higher weight than those in the transition and edge areas. The overall prediction error... The calculation formula is: in: This represents the total number of measurement points on the surface. Representing the Surface mechanical prediction values ​​at each measurement point Representing the The actual pressure signal values ​​collected at each measurement point Representing the The weight coefficients of each measurement point are used to determine whether the overall prediction error is less than the preset allowable error range. The allowable error range is set to five percent of the average amplitude of the actual collected pressure signal. When the overall prediction error is greater than or equal to the preset allowable error range, the gradient descent algorithm is used to adjust the activation level parameters of each muscle fiber bundle inside the gastrocnemius muscle biomechanical simulation model.

[0046] Optionally, the gradient descent algorithm iteratively updates the gradient direction and magnitude of the activation level parameters of each myofibril bundle based on the overall prediction error, with the update rule being: in: Representing the During the nth iteration Activation level parameters of strip muscle fiber bundles Represents the learning rate. Represents the overall prediction error for the first The partial derivatives of the activation level parameters of the muscle fiber bundles are calculated in each iteration using the adjoint variable method or the finite difference method. The forward calculation, error calculation, and parameter adjustment steps are repeated until the overall prediction error is less than the preset allowable error range.

[0047] It is understandable that when the overall prediction error is less than the preset allowable error range, the activation level parameters of each muscle fiber bundle inside the current gastrocnemius biomechanical simulation model are locked. The stress tensor data of each muscle fiber bundle in three-dimensional space is extracted from the locked gastrocnemius biomechanical simulation model to form stress field data. The stress tensor data is obtained by querying the finite element model result database or directly reading the solver output file. The stress tensor data of each muscle fiber bundle corresponds to the stress state at the Gaussian integration point of the finite element element occupied by the muscle fiber bundle. The stress state is represented in the form of a second-order tensor.

[0048] In some embodiments, referring to Table 1, a one-dimensional line integral is performed on the stress tensor data along the direction of the muscle fiber bundle within its spatial domain for each muscle fiber bundle. The direction of the muscle fiber bundle is defined by its geometric central axis, which is represented by a smooth curve formed by interpolation of a series of spatial coordinate points. The integral calculation is performed along this curve, and the result of the integral calculation is used as the instantaneous tension value of the muscle fiber bundle in the current motion state. The instantaneous tension values ​​of all muscle fiber bundles are collected to form an instantaneous tension distribution that reflects the distribution of tension magnitude within the gastrocnemius muscle. The instantaneous tension distribution is stored in the form of a data list. Each record in Table 1 contains a muscle fiber bundle identifier and the corresponding instantaneous tension value.

[0049] Table 1: Comparison of Predicted and Actual Pressure Values ​​at Surface Measurement Points As can be understood, Table 1 shows the pressure value comparison of some representative surface measurement points in a certain iteration step. The absolute difference will be used to calculate the weighted overall prediction error. The gradient descent algorithm continues to update iteratively until the prediction trend of all measurement points is consistent with the actual data trend and the overall prediction error is lower than the threshold. Locking the activation level parameter means fixing a set of parameter values ​​obtained in the current iteration as the optimal solution and using it for subsequent instantaneous tension distribution calculation. The extraction of stress tensor data and the calculation of line integral are automatically completed in batches in the model post-processing module. The instantaneous tension distribution serves as the input data for subsequent calculation of elastic potential energy storage and release efficiency.

[0050] See Figure 4In the overall dynamic response index analysis of the gastrocnemius muscle rebound force characteristics assessment, the time series plot fully presents the functional response characteristics of the gastrocnemius muscle within a typical contraction-relaxation cycle. Specifically, the plot uses time (s) as the horizontal axis and the overall dynamic response index (0-100) as the vertical axis, clearly dividing it into three functional stages: Initiation stage (red area, 0-2s): Both the raw data (gray curve) and the smoothed curve (orange curve) show a rapid rise followed by a fall, reflecting the process of the gastrocnemius muscle being activated by nerves from a resting state, rapid recruitment of muscle fibers, and generation of tension. The peak index is close to 80, indicating the explosive force level of the initial contraction. Stable contraction stage (green area, 2-6s): The curve enters the second cycle, and the upward slope and peak height (approximately 77) are similar to those of the initiation stage, indicating that the muscle fibers maintain a stable active contraction state under continuous nerve drive, with good consistency in tension output, reflecting the gastrocnemius muscle's sustained work capacity. Relaxation rebound stage (blue area, 6-10s): The curve completes the third cycle, with a peak of approximately 78, and then rapidly falls back to the baseline. This stage reflects the process by which muscle fibers, after active relaxation, rebound force is generated through the elastic recoil of tendons and extramuscular connective tissue. The rate of decrease in the index and the recovery speed from the baseline are directly related to the elastic potential energy release efficiency of the gastrocnemius muscle. The smooth curve (orange) in the figure, through noise reduction processing of the original data (gray), more clearly reveals the dynamic trends of each stage, providing key time-domain characteristic basis for subsequent inverse solving of muscle fiber tension distribution and calculation of elastic potential energy storage and release efficiency using biomechanical simulation models.

[0051] In one embodiment of the invention, for the instantaneous tension distribution curve of each muscle fiber bundle changing over time, the tension rise and tension fall segments within each movement cycle are identified. The area under the tension rise segment curve is calculated, representing the elastic potential energy stored by the muscle fiber bundle in each movement cycle. The area under the tension fall segment curve is calculated, representing the elastic potential energy released by the muscle fiber bundle in each movement cycle. The ratio of released elastic potential energy to stored elastic potential energy is defined as the single-cycle efficiency of the muscle fiber bundle. The single-cycle efficiency of all muscle fiber bundles is weighted and averaged according to their cross-sectional areas to obtain the overall elastic potential energy storage and release efficiency of the gastrocnemius muscle. A gastrocnemius muscle function level threshold database is established in advance by testing a large number of healthy individuals and individuals with different degrees of muscle dysfunction. This database divides the numerical range of elastic potential energy storage and release efficiency into five levels: excellent, good, moderate, impaired, and severely impaired. The calculated values ​​of the overall elastic potential energy storage and release efficiency of the gastrocnemius muscle of the subjects are matched with the ranges in this database. Output the matched functional level labels, and generate a structured evaluation report that includes the contribution of each feature analysis block and the main limiting factors.

[0052] In specific implementation, it involves calculating the elastic potential energy storage and release efficiency of the gastrocnemius muscle and determining the rebound force characteristic level. For the instantaneous tension distribution of each muscle fiber bundle over time, the tension rise segment and tension fall segment within each movement cycle are identified. The movement cycle is automatically divided according to the periodic characteristics of the plantar pressure signal or position signal. The tension rise segment is defined as the curve segment from the instantaneous tension minimum value to the maximum value within the cycle, and the tension fall segment is defined as the curve segment from the instantaneous tension maximum value within the cycle to the minimum value of the next cycle. The curve identification is completed based on the algorithm for finding local extreme points. The curve of instantaneous tension distribution over time is formed by connecting the instantaneous tension values ​​at each sampling time point calculated by the embodiment.

[0053] In some embodiments, the area under the curve during the tension rise segment is calculated to characterize the elastic potential energy stored in the muscle fiber bundle during each movement cycle, and the area under the curve during the tension fall segment is calculated to characterize the elastic potential energy released by the muscle fiber bundle during each movement cycle. The area under the curve is calculated using a numerical integration method, performing trapezoidal integral or Simpson integral on the discrete instantaneous tension data within the tension rise and tension fall segments. The ratio of released elastic potential energy to stored elastic potential energy is defined as the single-cycle efficiency of the muscle fiber bundle. The formula for calculating the single-cycle efficiency is as follows: in: Representing the Single-cycle efficiency of a strip muscle fiber bundle. Representing the The elastic potential energy released by a muscle fiber bundle during one movement cycle, i.e., the area under the tension decrease curve. Representing the The elastic potential energy stored in a bundle of muscle fibers within the same movement cycle is the area under the curve during the tension rise segment.

[0054] Optionally, the overall elastic potential energy storage and release efficiency of the gastrocnemius muscle can be obtained by weighting the efficiency of each cycle of all muscle fiber bundles according to their cross-sectional area. The weighted average formula is as follows: in: Represents the overall elastic potential energy storage and release efficiency of the gastrocnemius muscle. Represents the total number of muscle fiber bundles. Representing the The cross-sectional area of ​​the muscle fiber bundles is obtained from the geometric parameters of the gastrocnemius muscle biomechanical simulation model or muscle morphology measurements based on medical images. The weighted averaging process is performed in a dedicated efficiency calculation module.

[0055] Understandably, a gastrocnemius muscle function level threshold database is established in advance by testing a large number of healthy people and people with different degrees of muscle dysfunction. During the testing process, signals are collected from healthy people and people with muscle dysfunction when performing the same standardized exercise protocol, and their elastic potential energy storage and release efficiency is calculated. The database divides the numerical range of elastic potential energy storage and release efficiency into five levels: excellent, good, moderate, impaired, and severely impaired. Each level corresponds to a continuous efficiency value range. The boundary of the range is determined by statistical analysis of the efficiency values ​​of the test population combined with the interpretation of clinical assessment experts. The calculated value of the overall elastic potential energy storage and release efficiency of the gastrocnemius muscle of the subjects is matched with the range in the gastrocnemius muscle function level threshold database.

[0056] In some embodiments, the matched functional level labels are output as the final evaluation result. The functional level labels are presented in text form. At the same time, a structured evaluation report containing the contribution of each feature analysis block and the main limiting factors is generated. The contribution of each feature analysis block is calculated by analyzing the efficiency value of the myofiber bundles contained in each feature analysis block and the proportion of their cross-sectional area to the whole. The main limiting factors are determined by identifying the feature analysis blocks with the lowest contribution and the groups of myofiber bundles with individual cycle efficiencies significantly lower than the average level. The structured evaluation report is automatically generated by filling in the data using a predefined XML or JSON format template.

[0057] It is understandable that the numerical matching process of elastic potential energy storage and release efficiency is achieved by querying the threshold table in the database. The threshold table stores the lower and upper limits of efficiency values ​​corresponding to five levels. The matching logic is to determine which level range the subject's efficiency value falls into. The gastrocnemius muscle function level threshold database is updated regularly based on newly accumulated test data to optimize the accuracy of level classification. The structured assessment report includes quantitative data and qualitative description. The quantitative data section lists the elastic potential energy storage and release efficiency and the percentage contribution of each feature analysis block. The qualitative description section outputs prompt text based on the analysis of the main limiting factors. The assessment report is presented through a display device or printed out.

[0058] See Figure 5In the quantitative analysis of the elastic potential energy storage and release efficiency of the gastrocnemius muscle, the tension-time curve visually presents the mechanical behavior of the muscle fiber bundle within a single movement cycle. Specifically, the red curve and filled area represent the tension rise segment (energy storage phase), corresponding to the process of the muscle fiber bundle actively contracting from the minimum tension value to the maximum tension value within the cycle. The area under this curve is calculated using the trapezoidal integral or Simpson integral, and its value directly represents the total elastic potential energy stored by the muscle fiber bundle within this cycle. The blue curve and filled area represent the tension fall segment (energy release phase), corresponding to the process of the muscle fiber bundle relaxing from the maximum tension value to the minimum tension value of the next cycle. The area under this curve represents the total elastic potential energy released by the muscle fiber bundle within this cycle. The periodic fluctuations of the two curves reflect the mechanical response characteristics of the gastrocnemius muscle in a continuous contraction-relaxation cycle. The peak height, rise / fall slope, and fluctuation frequency of the curves can be further used to evaluate the contraction rate and elastic stiffness of the muscle fiber bundle. By calculating the ratio of the area of ​​the energy release phase to the area of ​​the energy storage phase, the single-cycle efficiency of a single muscle fiber bundle can be obtained. Then, by combining the weighted average of the cross-sectional areas of all muscle fiber bundles, the overall elastic potential energy storage and release efficiency of the gastrocnemius muscle can be obtained, providing a core quantitative indicator for subsequent muscle function level classification.

[0059] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for detecting muscle function to evaluate the rebound force characteristics of the gastrocnemius muscle, characterized in that, The method includes: The raw physiological and physical signals generated by the gastrocnemius muscle of the subject under specific exercise conditions were collected and the signals were decomposed to separate the active component characterizing the active contraction characteristics of the gastrocnemius muscle and the passive component characterizing the passive dynamic characteristics of the gastrocnemius muscle. The separated active and passive components are temporally aligned and feature-fused to construct a composite feature vector. Based on the anatomical structure and muscle fiber orientation of the gastrocnemius muscle, the corresponding body surface area is virtually divided into multiple hierarchical feature analysis blocks, and a weight coefficient is assigned to each feature analysis block. The state of each feature analysis block is assigned using the fused composite feature vector, and the overall dynamic response index of the gastrocnemius muscle is calculated based on the weight coefficient of each feature analysis block. Based on the overall dynamic response index, and combined with the pre-established gastrocnemius muscle biomechanical simulation model, the instantaneous tension distribution of muscle fiber bundles at different depths inside the gastrocnemius muscle under the specific motion state is solved in reverse. Based on the instantaneous tension distribution, the inflection point and periodic peak of tension change are identified, and then the elastic potential energy storage and release efficiency of the gastrocnemius muscle in the continuous contraction and relaxation cycle is calculated. The elastic potential energy storage and release efficiency is compared with multiple preset gastrocnemius muscle function level thresholds to determine the rebound force characteristic level of the subject's gastrocnemius muscle, thus completing the detection and evaluation of gastrocnemius muscle function.

2. The method for detecting muscle function to evaluate the rebound force characteristics of the gastrocnemius muscle as described in claim 1, characterized in that, The process involves signal decomposition to separate the active component characterizing the active contraction characteristics of the gastrocnemius muscle and the passive component characterizing the passive dynamic characteristics of the gastrocnemius muscle. Specifically, this includes: The original physiological signals include at least electromyographic signals, and the original physical signals include at least pressure signals and position signals; After performing multi-scale wavelet transform on the electromyographic signal to filter out electrocardiographic interference and motion artifacts, the neural activation sequence representing the recruitment and firing frequency of motor units was extracted as the active component. The pressure signal and the position signal are jointly analyzed. Baseline drift is removed by high-pass filtering. Then, the first and second derivatives of the signals are calculated to extract physical parameters that reflect the deformation rate, deformation acceleration and ground reaction force of the gastrocnemius muscle. The physical parameters are combined into passive components. A time delay correction relationship is established between the active and passive components. The conduction time from neural activation to mechanical response is determined through cross-correlation analysis. Based on this, the active component is time-adjusted in advance to complete the time-domain alignment.

3. The method for detecting muscle function to evaluate the rebound force characteristics of the gastrocnemius muscle as described in claim 2, characterized in that, The construction of the composite feature vector further includes: The time-aligned neural activation sequence is spliced ​​with the physical parameter sequence to form the initial fusion sequence; Principal component analysis is performed on the initial fusion sequence, and the first few principal components whose contribution rates exceed a preset threshold are retained to form a dimensionality-reduced feature subset. The principal components in the feature subset are respectively compared with the baseline features extracted from the gastrocnemius muscle signal of the subject at rest to obtain the dynamic feature increment after removing individual static differences. The dynamic feature increments are normalized, and the final output is the composite feature vector. The composite feature vector can simultaneously reflect the nerve activation state and mechanical response of the gastrocnemius muscle.

4. The method for detecting muscle function to evaluate the rebound force characteristics of the gastrocnemius muscle as described in claim 1, characterized in that, The process of virtually dividing the body surface region corresponding to the gastrocnemius muscle into multiple hierarchical feature analysis blocks and assigning weight coefficients to each feature analysis block includes: Obtain ultrasound or magnetic resonance images of the gastrocnemius muscle region of the subject, and mark the boundary between the medial and lateral heads of the gastrocnemius muscle, the location of the muscle belly, and the Achilles tendon junction area on the images; Based on the annotation information, the entire gastrocnemius muscle projection area is divided into three levels: the core area, the transition area, and the edge area. Within the core area, it is further subdivided into longitudinal muscle fiber-dominated areas and oblique muscle fiber-dominated areas based on the direction of the muscle bundles. Each segmented feature analysis block is assigned a weight coefficient, with the core area having a higher weight coefficient than the transition area, the transition area having a higher weight coefficient than the edge area, and the longitudinal muscle fiber dominant area having a higher weight coefficient than the oblique muscle fiber dominant area.

5. A method for detecting muscle function to evaluate the rebound force characteristics of the gastrocnemius muscle as described in claim 4, characterized in that, The calculation of the overall dynamic response index of the gastrocnemius muscle based on the weight coefficient of each feature analysis block includes: Each feature element in the composite feature vector is mapped and assigned according to its physiological meaning and the functional characteristics of the feature analysis block to obtain the feature value of each feature analysis block. The weighted contribution value of each feature analysis block is obtained by multiplying the feature value of each feature analysis block by the weight coefficient of the feature analysis block. The weighted contribution values ​​of all feature analysis blocks are superimposed linearly or nonlinearly, and the formula or model parameters of the operation are obtained by training and calibration on test data of the standard population in the early stage. The result of the superposition operation is scaled and mapped to the range of zero to one hundred, and the output is the overall dynamic response index.

6. A method for detecting muscle function to evaluate the rebound force characteristics of the gastrocnemius muscle as described in claim 5, characterized in that, The reverse calculation of the instantaneous tension distribution of muscle fiber bundles at different depths within the gastrocnemius muscle under a specific motion state includes: The gastrocnemius muscle biomechanical simulation model is a multi-scale finite element model that includes the epimysium, perimysium, muscle fibers, series elastic elements, and parallel elastic elements. The overall dynamic response index is used as the overall driving input of the gastrocnemius muscle biomechanical simulation model. In the aforementioned gastrocnemius muscle biomechanical simulation model, differentiated material property parameters and initial preload are set for muscle fiber bundles at different depths; The gastrocnemius muscle biomechanical simulation model is run for forward calculation. The surface mechanical prediction value output by the model is compared with the actual collected pressure signal. The activation level parameters of the muscle fiber bundles inside the model are adjusted by iterative optimization algorithm until the error between the predicted value and the actual value is less than the allowable range. The stress field data of each muscle fiber bundle within the optimized model are extracted and spatially integrated to obtain the instantaneous tension distribution of each muscle fiber bundle.

7. A method for detecting muscle function to evaluate the rebound force characteristics of the gastrocnemius muscle as described in claim 6, characterized in that, The gastrocnemius muscle biomechanical simulation model is run for forward calculation. The surface mechanical prediction values ​​output by the model are compared with the actual collected pressure signals. The activation level parameters of the muscle fiber bundles inside the model are adjusted through an iterative optimization algorithm until the error between the predicted value and the actual value is less than the allowable range. This includes: The overall dynamic response index is input into the gastrocnemius muscle biomechanical simulation model, and the forward calculation process of the model is executed to obtain the theoretical pressure value of each measurement point on the surface of the gastrocnemius muscle, which is used as the surface mechanical prediction value. Simultaneously acquire the pressure signal values ​​of each corresponding measurement point that are actually collected; Calculate the absolute difference between the predicted surface mechanical value and the actual collected pressure signal value at each measurement point; The total prediction error is obtained by summing the weighted absolute differences of all measurement points. Determine whether the overall prediction error is less than a preset allowable error range; When the overall prediction error is greater than or equal to the preset allowable error range, the gradient descent algorithm is used to adjust the activation level parameters of each muscle fiber bundle inside the gastrocnemius muscle biomechanical simulation model. Repeat the forward calculation, error calculation, and parameter adjustment steps until the overall prediction error is less than the preset allowable error range.

8. A method for detecting muscle function to evaluate the rebound force characteristics of the gastrocnemius muscle as described in claim 6, characterized in that, The step of extracting and spatially integrating the stress field data of each muscle fiber bundle within the optimized model to obtain the instantaneous tension distribution of each muscle fiber bundle includes: When the overall prediction error is less than the preset allowable error range, the activation level parameters of each muscle fiber bundle inside the current gastrocnemius muscle biomechanical simulation model are locked. The stress tensor data of each muscle fiber bundle in three-dimensional space are extracted from the locked gastrocnemius muscle biomechanical simulation model to form stress field data; For each muscle fiber bundle, the stress tensor data is calculated as a one-dimensional line integral along the direction of the muscle fiber bundle within its spatial domain. The integral calculation result is used as the instantaneous tension value of the muscle fiber bundle in the current motion state; The instantaneous tension values ​​of all muscle fiber bundles are collected to form the instantaneous tension distribution, which reflects the distribution of tension magnitude within the gastrocnemius muscle.

9. A method for detecting muscle function to evaluate the rebound force characteristics of the gastrocnemius muscle as described in claim 8, characterized in that, The calculation of the elastic potential energy storage and release efficiency of the gastrocnemius muscle in continuous contraction and relaxation cycles includes: For the curve of the instantaneous tension distribution of each muscle fiber bundle over time, identify the tension rise and tension fall segments within each movement cycle. Calculate the area under the tension rise curve, where the area represents the elastic potential energy stored in the muscle fiber bundle during each movement cycle; Calculate the area under the tension decrease curve, where the area represents the elastic potential energy released by the muscle fiber bundle in each movement cycle; The ratio of released elastic potential energy to stored elastic potential energy is defined as the single-cycle efficiency of the muscle fiber bundle. The overall elastic potential energy storage and release efficiency of the gastrocnemius muscle is obtained by weighting the individual cycle efficiency of all muscle fiber bundles according to their cross-sectional area.

10. A method for detecting muscle function to evaluate the rebound force characteristics of the gastrocnemius muscle as described in claim 1, characterized in that, The determination of the rebound force characteristics grade of the subject's gastrocnemius muscle includes: A threshold database of gastrocnemius muscle function levels was established by testing a large number of healthy people and people with different degrees of muscle dysfunction in advance. The database divides the numerical range of elastic potential energy storage and release efficiency into five levels: excellent, good, moderate, damaged, and severely damaged. The calculated values ​​of the elastic potential energy storage and release efficiency of the subject's gastrocnemius muscle as a whole are matched with the range in the gastrocnemius muscle functional level threshold database. Output the matched functional level labels, and generate a structured evaluation report that includes the contribution of each feature analysis block and the main limiting factors.

Images