Method and equipment for displaying muscle tissue morphology based on digital twin of levator scapulae muscle

Abstract

This invention provides a method and device for displaying muscle tissue morphology based on a digital twin of the levator scapulae muscle. The method includes acquiring static magnetic resonance imaging data of the target user's shoulder and neck region, multi-angle dynamic optical motion capture data, and surface electromyography (EMG) signals of the target muscle group to construct multimodal physiological data of the levator scapulae muscle. Based on the static magnetic resonance imaging data, a static anatomical base model of the levator scapulae muscle is constructed using image segmentation and 3D reconstruction techniques. According to the multi-angle dynamic optical motion capture data and EMG signals, the activation sequence and corresponding activation intensity of the target nerve driving the contraction of the levator scapulae muscle under different movements are calculated. Based on the activation sequence and corresponding activation intensity of the target nerve, the static anatomical base model of the levator scapulae muscle is embedded into a pre-trained digital twin model of the levator scapulae muscle to obtain a target digital twin of the levator scapulae muscle. The input motion command is input into the target digital twin of the levator scapulae muscle, and the muscle tissue morphology results are output and displayed.

Description

Technical Field

[0001] This disclosure relates to the field of medical technology, and in particular to a method and device for displaying muscle tissue morphology based on a digital twin of the levator scapulae muscle. Background Technology

[0002] The levator scapulae is a key muscle connecting the cervical spine and scapula, and its dysfunction is closely related to chronic neck pain, limited shoulder mobility, and postural dysphysiolysis. Traditional clinical assessment mainly relies on manual palpation, static medical imaging, and surface electromyography, but these methods are difficult to quantify the three-dimensional deformation, internal stress distribution, and neural activation patterns of the muscle during complex dynamic movements.

[0003] In relevant scenarios, biomechanical models based on human muscle modeling are used for analysis, but they lack personalized and dynamic descriptions of individual anatomical variations, real-time physiological states, and muscle activation strategies under specific tasks. This results in insufficient precision and targeting of interventions in rehabilitation program development, motor performance optimization, and ergonomic assessment. Summary of the Invention

[0004] In view of the aforementioned problems, this disclosure provides a method and device for displaying muscle tissue morphology based on a digital twin of the levator scapulae muscle, aiming to improve the accuracy of personalized and dynamic descriptions of individual anatomical variations, real-time physiological states, and muscle activation strategies under specific tasks, thereby improving the precision and targeting of intervention measures.

[0005] In conjunction with the first aspect of the present invention, an embodiment of the present invention provides a method for displaying muscle tissue morphology based on a digital twin of the levator scapulae muscle, comprising: acquiring static magnetic resonance imaging data of the shoulder and neck region of a target user, multi-angle dynamic optical motion capture data, and surface electromyography signals of a target muscle group; constructing multimodal physiological data of the levator scapulae muscle, wherein the target muscle group includes the levator scapulae muscle and synergistic muscle groups that coordinate with the movement of the levator scapulae muscle; and constructing a static anatomical base model of the levator scapulae muscle based on the static magnetic resonance imaging data, using image segmentation and three-dimensional reconstruction techniques, wherein the static anatomical base model of the levator scapulae muscle is based on the geometric morphology, attachment point, and fiber orientation of the levator scapulae muscle. The system constructs a first spatial relationship with adjacent bones and a second spatial relationship with adjacent ligaments of the levator scapulae. Based on the multi-angle dynamic optical motion capture data and the surface electromyography signals, it calculates the activation sequence and corresponding activation intensity of the target nerve driving the contraction of the levator scapulae under different movements. Based on the activation sequence and corresponding activation intensity of the target nerve, it embeds the static levator scapulae anatomical base model into a pre-trained levator scapulae digital twin model to obtain a target levator scapulae digital twin. In response to the acquired input motion command, it inputs the input motion command into the target levator scapulae digital twin and outputs and displays the muscle tissue morphology results of the levator scapulae.

[0006] In a preferred embodiment, the step of constructing a static levator scapulae anatomical base model based on the static magnetic resonance imaging data using image segmentation and 3D reconstruction techniques includes: automatically segmenting the image sequence corresponding to the static magnetic resonance imaging data using a deep convolutional neural network to identify the muscle-tendon boundaries of the levator scapulae muscle, the attachment points of the muscle to the transverse processes of the cervical vertebrae and the superior angle of the scapula, and obtaining identification result data; based on the identification result data, using a fiber tracing algorithm, reconstructing the spatial paths and curvatures of the main muscle fiber bundles within the muscle volume to obtain an initial muscle model, wherein the initial muscle model... The process involves refining a uniform entity into a collection of multiple representative muscle fiber bundles; generating a non-uniform tetrahedral mesh 3D model based on the initial muscle model; assigning anisotropic hyperelastic material initial property values ​​to each unit of each mesh in the tetrahedral mesh 3D model based on the anatomical location to which each unit belongs and the local diffuse anisotropy fraction along the muscle fiber direction; and adding a muscle fiber direction vector field to each unit of each mesh in the tetrahedral mesh 3D model based on the initial property values ​​to obtain the static levator scapulae anatomical base model.

[0007] In a preferred embodiment, assigning anisotropic initial values ​​of hyperelastic material properties to each cell of each grid in the tetrahedral mesh 3D model based on the anatomical location to which each cell belongs and the local diffuse anisotropy fraction along the muscle fiber direction includes: determining the anatomical location to which each cell belongs by comparing the geometric position, shape features, and relative relationship with surrounding structures of each cell in each grid of the tetrahedral mesh 3D model using a pre-constructed anatomical location database; and determining the anatomical location to which each cell belongs by reference physiological structure and reference motor function of the levator scapulae muscle. The muscle fiber direction of the anatomical location to which the unit belongs; based on the muscle fiber direction of the anatomical location to which each unit belongs, and combined with diffusion tensor imaging data, calculate the first diffusion coefficient along the muscle fiber direction and the second diffusion coefficient in the direction perpendicular to the muscle fiber direction for each unit; based on the first diffusion coefficient and the second diffusion coefficient corresponding to each unit, determine the degree of diffusion difference of the unit in different directions, and obtain the local diffusion anisotropy score; based on the anatomical location to which each unit belongs and the local diffusion anisotropy score, and referring to the mechanical property database of anisotropic hyperelastic materials, assign corresponding initial property values ​​to each unit.

[0008] In a preferred embodiment, the step of adding a muscle fiber direction vector field to each cell of each mesh in the tetrahedral mesh 3D model based on the initial attribute values ​​to obtain the static levator scapulae anatomical base model includes: obtaining the initial attribute values ​​of each cell in the tetrahedral mesh 3D model, analyzing their correlation with muscle fiber characteristics, and determining the distribution information of anisotropy and elastic modulus in different directions; based on the distribution information of each anisotropy and elastic modulus in different directions, formulating construction rules for the muscle fiber direction vector field, wherein the construction rules are used to specify the vector magnitude in the muscle fiber direction vector field, the association method of each anisotropy fraction, and the direction of the vector follows the muscle fiber direction vector field. Fiber physiological orientation information; according to the construction rules, for each unit in the tetrahedral mesh 3D model, with the corresponding initial attribute value as the target, add a corresponding muscle fiber direction vector field; according to the muscle fiber direction vector field, compare the tetrahedral mesh 3D model with the actual muscle tissue structure features; if the comparison result does not meet the preset conditions, readjust the initial attribute value, and repeat the steps of obtaining the initial attribute value of each unit in the tetrahedral mesh 3D model to comparing the tetrahedral mesh 3D model with the actual muscle tissue structure features until the comparison result meets the preset conditions, thus obtaining the static levator scapulae anatomical base model.

[0009] In a preferred embodiment, the step of calculating the activation sequence and corresponding activation intensity of the target nerve driving the contraction of the levator scapulae muscle under different movements based on the multi-angle dynamic optical motion capture data and the surface electromyography (EMG) signal includes: inputting the multi-angle dynamic optical motion capture data into an inverse kinematics solver based on the static levator scapulae anatomical base model to calculate the predicted length change curve of the levator scapulae muscle between the attachment points of the cervical spine and the scapula; determining the length change rate of the levator scapulae muscle based on the predicted length change curve; inputting the electrical signal sequence corresponding to the surface EMG signal into a potential convolution model based on double Hilbert transform to synthesize a theoretical high-density EMG of the corresponding muscle on the skin surface; estimating the activation sequence of the nerve activation driving the contraction of the levator scapulae muscle based on the theoretical high-density EMG, the predicted length change curve, and the length change rate; and determining the activation intensity of the target nerve driving the contraction of the levator scapulae muscle based on the length change rate.

[0010] In a preferred embodiment, estimating the activation sequence of the neural activation driving the contraction of the levator scapulae muscle based on the theoretical high-density electromyography (EDM), the predicted length change curve, and the length change rate includes: determining the predicted length value of the levator scapulae muscle at different times based on the predicted length change curve; establishing a relationship model between length and muscle contraction state by combining muscle physiological characteristics, analyzing the degree of contraction of the levator scapulae muscle at each time, and obtaining a length feature sequence reflecting the dynamic changes in the contraction of the levator scapulae muscle; calculating the length change amplitude of the levator scapulae muscle per unit time based on the length change rate; determining the acceleration or deceleration phase of the levator scapulae muscle contraction based on the length change amplitude and the length feature sequence, and obtaining the length change rate information of the levator scapulae muscle contraction; and estimating the activation sequence of the neural activation driving the contraction of the levator scapulae muscle based on the dimensionally unified basic EMG characteristics represented by the theoretical high-density EMG, the length feature sequence, and the length change rate information.

[0011] In a preferred embodiment, the step of embedding the static levator scapulae anatomical base model into a pre-trained levator scapulae digital twin model based on the activation sequence and corresponding activation intensity of the target nerve to obtain a target levator scapulae digital twin includes: fine-tuning the attention weights of the encoder in the pre-trained levator scapulae digital twin model based on the activation sequence and corresponding activation intensity of the target nerve, wherein the attention weights are the weights for fusing target features in multiple parallel branches of the encoder, and each of the multiple parallel branches performs feature extraction based on the input features to obtain the target features; wherein the decoder in the pre-trained levator scapulae digital twin model and the encoder... The encoder's fusion layer is used to fuse the target features of the multiple parallel branches according to the attention weights to form a latent space representation; the decoder is used to take the latent space representation as input and output the predicted three-dimensional deformation field data and predicted muscle stress-strain distribution data of the levator scapulae muscle; the static levator scapulae muscle anatomical base model is embedded downstream of the decoder's output to obtain the target levator scapulae muscle digital twin; wherein, the static levator scapulae muscle anatomical base model is used to perform muscle geometry transformation and tissue construction according to the predicted three-dimensional deformation field data and the predicted muscle stress-strain distribution data, and output and display the muscle tissue morphology results of the levator scapulae muscle.

[0012] In a preferred embodiment, acquiring static magnetic resonance imaging data of the target user's shoulder and neck region, multi-angle dynamic optical motion capture data, and surface electromyography (EMG) signals of the target muscle group includes: using a magnetic resonance diffusion tensor imaging sequence to acquire the conventional anatomical structure of the shoulder and neck region and extract the orientation and diffusion anisotropy fraction information of the levator scapulae muscle fibers to obtain an initial medical image; performing semantic segmentation of the tissue structure and generating a three-dimensional mesh on the initial medical image to obtain the static magnetic resonance imaging data; using a marker cluster to capture the three-dimensional motion trajectory of the cervical spine to thoracic spine, clavicle, and scapula when the target user performs the full range of movements of the levator scapulae muscle to obtain motion capture data; and performing skeleton binding and coordinate system unification on the motion capture data to obtain the multi-angle dynamic optical motion capture data; acquiring initial EMG signals by covering the upper trapezius muscle, the surface projection area of ​​the levator scapulae muscle, and the rhomboid muscle region with an electrode array; and performing bandpass filtering, rectification, and blind source separation algorithm on the EMG signals to decompose the motor unit firing sequence for the levator scapulae muscle to obtain the surface EMG signals of the target muscle group.

[0013] In conjunction with a second aspect of the present invention, an embodiment of the present invention provides a muscle tissue morphology display device based on a digital twin of the levator scapulae muscle, comprising: an acquisition module configured to acquire static magnetic resonance imaging data of the shoulder and neck region of a target user, multi-angle dynamic optical motion capture data, and surface electromyography signals of a target muscle group, and construct multimodal physiological data of the levator scapulae muscle, wherein the target muscle group includes the levator scapulae muscle and synergistic muscle groups that coordinate the movement of the levator scapulae muscle; and a construction module configured to construct a static anatomical base model of the levator scapulae muscle based on the static magnetic resonance imaging data, using image segmentation and three-dimensional reconstruction techniques, wherein the static anatomical base model of the levator scapulae muscle is based on the geometric shape of the levator scapulae muscle, the attachment point of the levator scapulae muscle, the direction of the muscle fibers, and the relationship between the levator scapulae muscle and adjacent bones. The system is constructed based on the first spatial relationship and the second spatial relationship between the levator scapulae muscle and the adjacent ligaments; the calculation module is configured to calculate the activation sequence and corresponding activation intensity of the target nerve driving the contraction of the levator scapulae muscle under different movements based on the multi-angle dynamic optical motion capture data and the surface electromyography signal; the embedding module is configured to embed the static levator scapulae muscle anatomical base model into the pre-trained levator scapulae muscle digital twin model based on the activation sequence and corresponding activation intensity of the target nerve to obtain the target levator scapulae muscle digital twin; the response output module is configured to respond to the acquired input motion command, input the input motion command into the target levator scapulae muscle digital twin, and output and display the muscle tissue morphology results of the levator scapulae muscle.

[0014] In conjunction with a third aspect of the present invention, an electronic device is provided in an embodiment of the present invention, comprising: a memory having a computer program stored thereon; and a processor for executing the computer program in the memory to implement the steps of the method described in any of the first aspects.

[0015] This disclosure, through the aforementioned technical solution, achieves at least the following effective results: By integrating static anatomical structures and dynamic motion data through multimodal data fusion and digital twin modeling, the constructed digital twin can quantify individual muscle geometry, spatial relationships, and neural activation patterns, overcoming the limitations of traditional models in adapting to anatomical variations. Based on closed-loop feedback of real-time neural activation parameters and movement commands, the model can dynamically predict muscle morphological changes and mechanical responses, supporting real-time adjustments to intervention strategies and avoiding intervention failures due to individual differences or task changes. By analyzing the interaction mechanism of the muscle-nerve-skeleton system, key pathological links (such as trigger points and synergistic muscle imbalances) can be located, improving rehabilitation efficiency and motor performance while reducing the risk of over-intervention or misoperation. It achieves precise dynamic characterization of individual levator scapulae anatomical features, real-time physiological state, and task-related muscle activation strategies.

[0016] Other features and advantages of this disclosure will be described in detail in the following detailed description section. Attached Figure Description

[0017] The accompanying drawings are provided to further illustrate the present disclosure and form part of the specification. They are used together with the following detailed description to explain the present disclosure, but do not constitute a limitation thereof. In the drawings: Figure 1 This is a schematic diagram of the execution flow of a muscle tissue morphology display method based on a digital twin of the levator scapulae muscle provided in an embodiment of the present invention.

[0018] Figure 2 This is one implementation provided by an embodiment of the present invention. Figure 1 A schematic diagram of the execution flow of step S12.

[0019] Figure 3 This is one implementation provided by an embodiment of the present invention. Figure 1 A schematic diagram of the execution flow of step S13.

[0020] Figure 4 This is a block diagram of a muscle tissue morphology display device based on a digital twin of the levator scapulae muscle, provided in an embodiment of the present invention.

[0021] Figure 5 This is a schematic diagram of exemplary hardware and software components of a muscle tissue morphology display device based on a digital twin of the levator scapulae muscle, provided in an embodiment of the present invention. Detailed Implementation

[0022] 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.

[0023] The specific embodiments of this disclosure will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit this disclosure.

[0024] The acquisition, storage, use, and processing of data in this application comply with relevant laws and regulations, and the acquisition, storage, use, and processing of user information in this application's implementation method have been authorized and agreed upon by the customer.

[0025] This invention provides a method for displaying muscle tissue morphology based on a digital twin of the levator scapulae muscle. (See also...) Figure 1 As shown, the method includes: In step S11, static magnetic resonance imaging data of the target user's shoulder and neck region, multi-angle dynamic optical motion capture data, and surface electromyography signals of the target muscle group are acquired to construct multimodal levator scapulae physiological data. The target muscle group includes the levator scapulae and the synergistic muscle group that works in conjunction with the movement of the levator scapulae. Among them, static magnetic resonance imaging data are high-resolution anatomical images of the shoulder and neck region acquired through magnetic resonance imaging, which can clearly show the geometry and spatial relationships of soft tissues such as muscles, bones, and ligaments. Multi-angle dynamic optical motion capture data uses optical sensors (such as camera arrays) to collect the three-dimensional coordinates, angles, and motion trajectories of joints during shoulder and neck region movement, reflecting the dynamic displacement of bones. Surface electromyography (EMG) signals are collected through electrode patches to capture the weak electrical signals generated during muscle contraction, reflecting the activation timing and intensity of target muscle groups (such as the levator scapulae and its synergistic muscles).

[0026] In this embodiment, the static anatomical structure of the shoulder and neck region is acquired through magnetic resonance imaging (MRI) scanning. An optical motion capture system (such as 3DAT technology) uses a multi-camera array to record the shoulder joint movement trajectory and simultaneously acquires sEMG signals of the levator scapulae and synergistic muscle groups. During data fusion, timestamps are used to align static images and dynamic data to ensure the spatiotemporal consistency between the anatomical base model and the motion data. For example, during cervical lateral flexion, the motion capture system records the scapular rotation angle, and surface electromyography (SEMG) signals synchronously reflect the temporal differences in activation of the levator scapulae and trapezius muscles.

[0027] In step S12, based on the static magnetic resonance imaging data, a static levator scapulae anatomical base model is constructed using image segmentation and three-dimensional reconstruction techniques. The static levator scapulae anatomical base model is constructed based on the geometric shape of the levator scapulae, the attachment point of the levator scapulae, the direction of the muscle fibers, the first spatial relationship between the levator scapulae and adjacent bones, and the second spatial relationship between the levator scapulae and adjacent ligaments. In this embodiment, a segmentation algorithm based on active contours is used to process static magnetic resonance imaging data. A scale-adjustable model driven by robust statistics overcomes grayscale inhomogeneity and accurately segments the contour of the levator scapulae muscle. Combined with 3D reconstruction technology, the segmented 2D contour is converted into a 3D mesh model, quantifying the muscle fiber orientation (e.g., the sagittal angle from the transverse processes of C1-C4 to the superior angle of the scapula), attachment point coordinates (e.g., anatomical landmarks on the medial border of the scapula), and spatial relationship with the cervical spine and scapula (e.g., the distance between the muscle belly and the coracoid ligament).

[0028] In step S13, based on the multi-angle dynamic optical motion capture data and the surface electromyography signal, the activation sequence and corresponding activation intensity of the target nerve driving the levator scapulae muscle to contract under different movements are calculated. In this embodiment, shoulder joint kinematic parameters (such as angular velocity and acceleration) are extracted based on dynamic motion capture data. Combined with the time-frequency characteristics of sEMG signals (such as the decreasing trend of average power frequency (MPF), a deep learning algorithm (such as the PhysCap model) is used to calculate the neural activation sequence driving the contraction of the levator scapulae muscle. For example, during shoulder abduction, the model identifies a high-frequency burst of sEMG signals in the initial stage of movement (0-0.5 seconds), corresponding to rapid activation of the C3-C5 nerve roots; the signal frequency decreases in the maintenance stage of movement (0.5-2 seconds), reflecting a weakening of neural activation intensity to avoid muscle fatigue.

[0029] In step S14, based on the activation sequence and corresponding activation intensity of the target nerve, the static levator scapulae anatomical base model is embedded into the pre-trained levator scapulae digital twin model to obtain the target levator scapulae digital twin. In this embodiment, a static anatomical base model is imported into a pre-trained digital twin framework for the levator scapulae muscle. This framework integrates a biomechanical model (such as the Hill muscle model) and a neural control model (such as a recurrent neural network). Based on the neural activation parameters calculated in step S13, the contractile force-length curve of the muscle in the digital twin is adjusted (e.g., optimizing the relationship between muscle fiber shortening speed and force) and the neural conduction delay (e.g., setting an electrode delay time of 10 ms) to ensure that the error between the model output and the actual motion capture data is less than 5%.

[0030] In step S15, in response to the acquired input motion command, the input motion command is input into the target levator scapulae digital twin, and the muscle tissue morphology result of the levator scapulae is output and displayed.

[0031] In this embodiment of the disclosure, when a user inputs a movement command (such as "rotate the cervical spine 30°"), the digital twin simulates the morphological changes of the levator scapulae muscle through finite element analysis based on the current neural activation state (such as the model parameters updated in step S14) and kinematic constraints (such as shoulder joint range of motion limitations). For example, when the cervical spine rotates to 25°, the model predicts that the length of the levator scapulae muscle belly shortens by 2 mm and the tendon tension increases by 15 N, and displays the dynamic stretching process of the muscle fibers in a three-dimensional visualization.

[0032] The aforementioned technical solution integrates static anatomical structures and dynamic motion data through multimodal data fusion and digital twin modeling. The constructed digital twin can quantify individual muscle geometry, spatial relationships, and neural activation patterns, overcoming the limitations of traditional models in adapting to anatomical variations. Based on closed-loop feedback of real-time neural activation parameters and movement commands, the model can dynamically predict muscle morphological changes and mechanical responses, supporting real-time adjustments to intervention strategies and avoiding intervention failures due to individual differences or task variations. By analyzing the interaction mechanisms of the muscle-nerve-skeleton system, key pathological links (such as trigger points and synergistic muscle imbalances) can be located, improving rehabilitation efficiency and motor performance while reducing the risk of over-intervention or misoperation. It achieves precise dynamic characterization of individual levator scapulae anatomical features, real-time physiological state, and task-related muscle activation strategies.

[0033] In a preferred embodiment, see Figure 2 As shown, in step S12, the construction of a static levator scapulae anatomical base model based on the static magnetic resonance imaging data, using image segmentation and three-dimensional reconstruction techniques, includes: In step S121, a deep convolutional neural network is used to automatically segment the image sequence corresponding to the static magnetic resonance imaging data, identify the muscle-tendon boundary of the levator scapulae muscle, the attachment point of the muscle to the transverse process of the cervical vertebra and the superior angle of the scapula, and obtain the identification result data. In this embodiment, a DCNN model with a U-Net architecture is used. Its encoder extracts multi-scale features of the image (such as muscle belly texture and attachment point highlight areas) through continuous convolution and downsampling, while the decoder restores spatial resolution through deconvolution and skip connections, ultimately outputting a pixel-level segmentation mask. During training, the model uses doctor-annotated MRI data as a supervision signal to optimize the cross-entropy loss function, enabling it to learn to distinguish the boundary between the levator scapulae muscle and adjacent tissues (such as the trapezius muscle and cervical fascia).

[0034] For example, in the cervical transverse process attachment area, the model uses a local contrast enhancement module to focus on gray-level abrupt changes at the bone-muscle junction, accurately identifying the attachment point coordinates. The DCNN model identifies gray-level gradient abrupt changes (low MRI signal in bone tissue, high signal in muscle) at the bone-muscle junction around the attachment point, and focuses on this area through an attention mechanism, with a segmentation error of less than 0.5 mm.

[0035] In step S122, based on the identification result data, a fiber tracing algorithm is used to reconstruct the spatial path and curvature of the main muscle fiber bundles within the muscle volume to obtain an initial muscle model. The initial muscle model is a set of multiple representative muscle fiber bundles that refine a uniform entity. Among them, the fiber tracing algorithm is based on the principle of diffusion tensor imaging. It reconstructs the spatial path of muscle fiber bundles by analyzing the diffusion directionality (anisotropy) of water molecules, and quantifies the orientation and curvature of muscle fibers.

[0036] In this embodiment, the volume of the levator scapulae muscle is extracted based on the segmentation results. A deterministic fiber tracing algorithm (such as the FACT algorithm) is used, starting from a seed point (such as the center of the muscle belly), and iteratively extending the fiber path along the local maximum diffusion direction until it encounters a region with low local diffusion anisotropy (such as a tendon) or terminates at a boundary. During the tracing process, the step size (0.5-2 mm) and angle threshold (30°-60°) are dynamically adjusted to balance tracing accuracy and integrity. The final generated initial model contains 500-1000 representative muscle fiber bundles, each bundle having its spatial path described by continuous point coordinates, and the curvature is calculated by the angle between the normal vectors of adjacent points.

[0037] For example, starting from the seed point at the center of the muscle belly, the algorithm tracks along the direction where the local diffusion anisotropy value is >0.2, and terminates at the attachment point because the local diffusion anisotropy value drops sharply (<0.1), generating a fiber bundle with a length of 15mm and a radius of curvature of 8mm, which accurately covers the real anatomical path.

[0038] In step S123, a non-uniform tetrahedral mesh three-dimensional model is generated based on the initial muscle model. Based on the anatomical location to which each unit of each mesh in the tetrahedral mesh three-dimensional model belongs and the local diffuse anisotropy fraction along the muscle fiber direction, an anisotropic hyperelastic material property initial value is assigned to each unit of each mesh in the tetrahedral mesh three-dimensional model. The local diffusion anisotropy fraction (FA) is an indicator (0-1) used to quantify the consistency of water molecule diffusion direction within muscle fibers. A higher FA value indicates a more ordered arrangement of muscle fibers, guiding the anisotropic distribution of material properties. The FA value ranges from 0 to 1, with FA=0 representing complete isotropy (e.g., liquids) and FA=1 representing complete anisotropy (e.g., highly ordered muscle fibers), guiding the distribution of material properties. Among them, hyperelastic material properties are material parameters used to describe the nonlinear stress-strain relationship of muscles under large deformation, including elastic modulus, Poisson's ratio, etc. Here, the local diffuse anisotropy fraction is dynamically adjusted to simulate the mechanical properties of real muscle fibers.

[0039] In this embodiment, the Delaunay tetrahedral subdivision algorithm is used to discretize the muscle volume into a non-uniform mesh. The mesh density is higher in the muscle belly (high strain region) (unit side length 0.5-1 mm) and lower in the tendon (low strain region) (2-3 mm). Material properties are assigned according to the anatomical location of the mesh unit (e.g., muscle belly, attachment point) and the local FA value: muscle belly units use high elastic modulus (10-50 kPa) and low anisotropy ratio (1.2-1.5) to simulate lateral contraction; attachment point units use low elastic modulus (1-5 kPa) and high anisotropy ratio (2-3) to simulate the tensile properties of the tendon. The FA value guides the property assignment through linear mapping (FA=0 → isotropic; FA=1 → perfectly anisotropic).

[0040] For example, the grid cell in the attachment point region has a side length of 1 mm, an FA value of 0.15, a distributed elastic modulus of 3 kPa, and an anisotropy ratio of 2.2; the adjacent muscle belly cell has a side length of 0.7 mm, an FA value of 0.35, a distributed elastic modulus of 20 kPa, and an anisotropy ratio of 1.3.

[0041] In step S124, based on the initial attribute values, a muscle fiber direction vector field is added to each cell of each mesh in the tetrahedral mesh 3D model to obtain the static levator scapulae anatomical base model.

[0042] Among them, the muscle fiber direction vector field is a set of directional vectors defined for each cell in the three-dimensional mesh model. The direction of the vector represents the physiological direction of the muscle fiber (such as the levator scapulae muscle pointing from the transverse process of the cervical vertebra to the superior angle of the scapula), and the magnitude reflects the density of local muscle fibers or the dominant mechanical direction.

[0043] In this embodiment, for each unit, the average direction vector of its internal muscle fiber bundles is calculated (determined by the eigenvector of the covariance matrix of all points within the bundle), and this vector is assigned to the unit's center point, forming a continuous direction vector field. This vector field is used to define the direction of muscle contraction in subsequent mechanical simulations: when simulating levator scapulae contraction, the unit generates active force along the vector direction, while only passive force is transmitted in the perpendicular direction. For example, in a cervical spine rotation task, the vector field guides the model to predict the synergistic effect of muscle belly shortening (active contraction) and tendon stretching (passive deformation). For example, the unit's direction vector forms a 12° angle with the long axis of the cervical transverse process, simulating the anatomical features of real muscle fibers attaching to the bone surface at an angle of 10°-15°.

[0044] The aforementioned technical solution achieves high-precision digital reconstruction of the levator scapulae muscle's anatomical structure through deep learning segmentation, fiber tracing, and non-uniform mesh modeling, solving the problem of poor adaptability of traditional models to individual anatomical variations (such as attachment point position shifts and differences in muscle fiber orientation). Constructing a static base model allows for the quantification of muscle geometry, spatial relationships, and material properties, improving the reliability of dynamic simulations (such as predicting muscle deformation under movement tasks). It can also improve the accuracy of levator scapulae trigger point localization in auxiliary diagnosis, optimize surgical path planning, and enhance the precision and safety of intervention measures.

[0045] In a preferred embodiment, step S123, assigning initial values ​​of anisotropic hyperelastic material properties to each cell of each grid in the tetrahedral mesh three-dimensional model based on the anatomical location to which each cell belongs and the local diffuse anisotropy fraction along the muscle fiber direction, includes: In step S1231, based on a pre-built anatomical site database, the geometric position, shape features, and relative relationship with the surrounding structures of each cell in the tetrahedral mesh 3D model are compared to determine the anatomical site to which each cell belongs. The anatomical location database contains a standardized dataset of the geometric location, shape features, and spatial relationships of the levator scapulae muscle and its adjacent structures (such as the transverse processes of the cervical vertebrae and the superior angle of the scapula), which is used for the anatomical classification of grid cells.

[0046] In this embodiment, based on a pre-constructed anatomical site database (containing geometric templates for regions such as the levator scapulae muscle belly, tendon, and attachment point), a spatial registration algorithm matches the coordinates, shape (e.g., aspect ratio), and adjacency (e.g., distance to the transverse process of the cervical vertebra) of mesh units with the database templates. For example, tendon units are classified as attachment point regions because their aspect ratio is >3 and their distance from the bone surface is <1mm; muscle belly units are classified as contraction regions because their volume percentage is >70% and their FA value is high. A weighted voting mechanism is used during matching to improve the accuracy of classification by combining geometric, positional, and topological features. For example, a unit with an aspect ratio of 2.5 and a distance of 0.8mm from the transverse process of the cervical vertebra is matched with the tendon-bone junction template in the database and classified as an attachment point region.

[0047] In step S1232, the direction of muscle fibers of the anatomical location to which each unit belongs is determined based on the reference physiological structure and reference motor function of the levator scapulae muscle. In this embodiment, referencing the physiological functions of the levator scapulae muscle (such as elevating the scapula and extending the cervical spine), and combining anatomical knowledge, the dominant direction of muscle fibers in each location is defined. For example, the direction of the muscle belly unit is along the long axis of the muscle (the angle between the muscle belly unit and the line connecting the superior angle of the scapula to the transverse process of the cervical vertebra is <15°); the direction of the attachment point unit is perpendicular to the bone surface (the angle between the attachment point unit and the normal vector of the transverse process of the cervical vertebra is <10°), to simulate the tensile strength of tendons. The direction definition is achieved by using principal component analysis (PCA) to reduce the dimensionality of the local mesh vertex coordinates and extracting the direction with the maximum variance as the muscle fiber direction.

[0048] In step S1233, based on the muscle fiber direction of the anatomical location to which each unit belongs, and combined with diffusion tensor imaging data, a first diffusion coefficient along the muscle fiber direction and a second diffusion coefficient in the direction perpendicular to the muscle fiber direction are calculated for each unit. Diffusion tensor imaging generates anisotropy fractions by measuring the diffusion directionality (diffusion tensor) of water molecules in tissue, which reflects the degree of orderliness in the arrangement of muscle fibers. The first diffusion coefficient and the second diffusion coefficient represent the diffusion rate of water molecules along the muscle fiber direction and perpendicular to the direction, respectively, and are used to quantify the degree of anisotropy.

[0049] In this embodiment, the diffusion tensor matrix (3×3 symmetric matrix) of the region corresponding to each grid cell is extracted from the DTI data, and the diffusion coefficient is obtained through eigenvalue decomposition. The first diffusion coefficient is the maximum diffusion rate along the muscle fiber direction, reflecting the rapid flow of water molecules within the muscle fiber; the second diffusion coefficient is the average diffusion rate in the vertical direction, reflecting the resistance of the intercellular matrix or muscle membrane. Trilinear interpolation is used to map the DTI voxel data to the center of the grid cell during calculation.

[0050] In step S1234, based on the first diffusion coefficient and the second diffusion coefficient corresponding to each unit, the degree of diffusion difference of the unit in different directions is determined to obtain the local diffusion anisotropy score. In this embodiment of the disclosure, it can be achieved through formulas The local diffusion anisotropy fraction is calculated to quantify the difference in diffusion rates in three directions. Here, λ1 is the first diffusion coefficient, λ2 is the coefficient of the second diffusion coefficient along the first direction, and λ3 is the coefficient of the second diffusion coefficient along the second direction. The first and second directions are opposite directions. It is the average of the coefficients in the first direction and the coefficients in the second direction.

[0051] For example, the muscle belly unit exhibits high anisotropy due to its first diffusion coefficient being much greater than the second diffusion coefficient (FA>0.4), simulating the orderly arrangement of muscle fibers; the attachment point unit exhibits low anisotropy due to its first diffusion coefficient being close to the second diffusion coefficient (FA<0.2), simulating the properties of the adhesive material of the tendon. The FA value range is mapped to the 0-1 interval through normalization.

[0052] In step S1235, based on the anatomical location to which each unit belongs and the local diffusion anisotropy fraction, and referring to the mechanical property database of anisotropic hyperelastic materials, corresponding initial property values ​​are assigned to each unit.

[0053] In this embodiment, based on the anatomical location (e.g., muscle belly, attachment point) and FA value (high / low anisotropy), corresponding Mooney-Rivlin model parameters (C10, C01, D1) are matched from a hyperelastic material database. For example, high FA muscle belly units are assigned C10=10kPa and C01=2kPa (high elasticity along the muscle fiber direction, low elasticity in the perpendicular direction) to simulate active contraction; low FA attachment point units are assigned C10=3kPa and C01=8kPa (equivalent elasticity in all directions, simulating tensile strength), while an anisotropy ratio (C10 / C01) is introduced to enhance direction dependence. Linear interpolation is used to handle the transition region between high and low FA values ​​during allocation.

[0054] The aforementioned technical solution achieves a high degree of personalization and biorealism in the material properties of the levator scapulae 3D model through precise anatomical location attribution, physiological definition of muscle fiber direction, and DTI-driven anisotropic material allocation. Material properties are deeply matched to the muscle's physiological role (e.g., high elasticity in the contraction zone, tensile strength at the attachment point), avoiding simulation errors caused by neglecting anatomical heterogeneity in traditional homogeneous models. FA-guided property allocation reflects individual differences in muscle fiber arrangement (e.g., athletes have more ordered muscle fibers), supporting adaptive adjustments of the model to different populations (e.g., patients, athletes). This enhances the accuracy and safety of intervention programs (e.g., rehabilitation training, surgical approaches).

[0055] In a preferred embodiment, step S124, which involves adding a muscle fiber direction vector field to each cell of each mesh in the tetrahedral mesh 3D model based on the initial attribute value to obtain the static levator scapulae anatomical base model, includes: In step S1241, the initial values ​​of the properties of each unit in the tetrahedral mesh three-dimensional model are obtained, the correlation with the characteristics of muscle fibers is analyzed, and the distribution information of anisotropy and elastic modulus in different directions is determined. In this embodiment, anisotropic core parameters are extracted from initial attribute values ​​(e.g., elastic modulus E1=15kPa, E2=3kPa, FA=0.7): the directional elasticity difference is quantified by E1 / E2=5, and the FA value reflects the orderly arrangement of muscle fibers. Further analysis of the spatial distribution of parameters reveals that, for example, muscle belly units exhibit strong anisotropy due to high FA (>0.6), with the elastic modulus dominating along the muscle fiber direction; attachment point units exhibit weak anisotropy due to low FA (<0.3), with the elastic modulus being similar in all directions. The distribution information is used to generate a continuous field through spatial interpolation (e.g., Kriging), avoiding simulation distortion caused by abrupt parameter changes between units.

[0056] In step S1242, based on the distribution information of each degree of anisotropy and elastic modulus in different directions, a construction rule for the muscle fiber direction vector field is formulated. The construction rule is used to specify the vector magnitude in the muscle fiber direction vector field, the association method of each anisotropy fraction, and the muscle fiber physiological orientation information followed by the direction of the vector. The degree of anisotropy is quantified by the FA value or the ratio of elastic modulus (such as E1 / E2), which is used to describe the difference in mechanical properties of a material in a certain direction (such as the direction of muscle fibers) and the perpendicular direction. High anisotropy indicates that the material is more easily deformed or subjected to load in a specific direction.

[0057] In this embodiment, the vector magnitude is positively correlated with the degree of anisotropy. High FA units have larger vector magnitudes (e.g., FA=0.7 corresponds to a magnitude of 0.8), while low FA units have smaller magnitudes (FA=0.2 corresponds to a magnitude of 0.3), reflecting muscle fiber density or mechanical dominance. The FA is associated by aligning the vector direction with the physiological orientation of the muscle fibers, and the directional deviation is adjusted by FA weighting (e.g., directional deviation <5° when FA=0.9, and allowable deviation <15° when FA=0.3). The physiological orientation information is based on the anatomically defined muscle fiber path (e.g., the levator scapulae muscle from the posterior tubercle of the transverse process of C1-C4 to the medial border of the superior angle of the scapula), and a smooth direction field is generated by B-spline curve fitting to avoid local directional abrupt changes.

[0058] In step S1243, according to the construction rules, for each unit in the tetrahedral mesh 3D model, a corresponding muscle fiber direction vector field is added with the corresponding initial attribute value as the target. In this embodiment, for each mesh element, a corresponding vector is matched from the construction rules based on its initial attribute values ​​(e.g., E1=12kPa, E2=4kPa, FA=0.65): The vector magnitude is calculated as follows: Magnitude = 0.7×FA + 0.1 = 0.7×0.65 + 0.1 = 0.555 (normalized to the 0-1 interval); Direction is determined by extracting the direction vector of the element's center point from a predefined muscle fiber direction field (e.g., along the line connecting the superior angle of the scapula to the transverse process of the cervical vertebra); Assignment is performed by combining the magnitude and direction to form the element's muscle fiber direction vector (e.g., vector (0.5, 0.3, 0.2) represents the component along the xyz direction). During addition, the continuity between the vector and adjacent elements needs to be checked, and local direction noise is eliminated through Laplacian smoothing.

[0059] In step S1244, the tetrahedral mesh three-dimensional model is compared with the actual muscle tissue structure features based on the muscle fiber direction vector field. In this embodiment, the generated vector field model is compared with the anatomical features of real muscles (such as the direction of muscle fibers in MRI images and the distribution of FA in DTI data), with a focus on verifying: directional consistency: the angle between the vector direction of more than 90% of the units in the model and the main diffusion direction of DTI (λ1 direction) is <15°; anisotropy matching: the average error between the model FA value and the DTI measurement value is <0.1 (e.g., model FA=0.68 vs. DTI FA=0.72); anatomical structure alignment: the vector field distribution of areas such as muscle belly, tendon, and attachment point is consistent with the anatomical atlas (e.g., the vector magnitude of the tendon area is <0.3 and the direction is perpendicular to the bone surface).

[0060] In step S1245, if the comparison result does not meet the preset conditions, the initial value of the attribute is readjusted, and the steps of obtaining the initial value of the attribute of each unit in the tetrahedral mesh three-dimensional model and comparing the tetrahedral mesh three-dimensional model with the actual muscle tissue structure features are repeated until the comparison result meets the preset conditions, and the static levator scapulae anatomical base model is obtained.

[0061] In this embodiment of the disclosure, if the comparison does not meet the preset conditions (such as direction error > 15° or FA error > 0.1), the initial values ​​of the attributes are adjusted in reverse: Direction error correction: If the model vector direction is offset, the vector field is regenerated by rotating the principal axis of elastic modulus (such as rotating the E1 direction by 5°); FA error correction: If the model FA value is too low, the E1 / E2 ratio of the corresponding unit is increased (such as from 5 to 6) or the DTI data interpolation weight is adjusted; Iteration termination condition: When the direction error of two consecutive comparisons is < 10° and the FA error is < 0.08, the iteration stops and the final anatomical base model is output.

[0062] For example, taking a mesh cell in the middle of the belly of the levator scapulae muscle as an example: Attribute analysis: This cell has E1=14kPa, E2=2.8kPa, and FA=0.8, showing strong anisotropy; Rule matching: Based on FA=0.8, the vector magnitude is 0.7×0.8+0.1=0.66, and the direction is extracted from the predefined field as the line connecting the superior angle of the scapula to the transverse process of the cervical vertebra (direction vector (0.7, 0.6,0.2)); Comparison and verification: Compared with DTI data, the model vector direction has an angle of 12° (<15°) with the λ1 direction, and the FA error is 0.03 (<0.1), which meets the conditions; Iterative adjustment: If the neighboring cell has a vector magnitude of 0.45 and a direction deviation of 18° due to FA=0.5, then adjust its E1 / E2 from 4 to 5, and after regenerating the vector field, the deviation is reduced to 10°, and the final model passes the verification.

[0063] The aforementioned technical solution achieves a high degree of biofidelity in the static levator scapulae model through the synergistic optimization of initial attribute values ​​and the muscle fiber orientation vector field. The vector field not only reflects the physiological orientation of the muscle fibers but also quantifies their mechanical anisotropy through initial attribute values, avoiding simulation distortion caused by the separation of orientation and attributes in traditional models. Through an iterative adjustment mechanism, the model can automatically correct initial parameter errors, ensuring a high degree of matching with the structure (such as the muscle belly-tendon transition zone) and function (such as direction-dependent elasticity) of real muscles; thus improving the accuracy of diagnosis and intervention programs for muscle-related diseases.

[0064] In a preferred embodiment, see Figure 3 As shown, in step S13, calculating the activation sequence and corresponding activation intensity of the target nerve driving the levator scapulae muscle contraction under different movements based on the multi-angle dynamic optical motion capture data and the surface electromyography signal includes: In step S131, the multi-angle dynamic optical motion capture data is input into the inverse kinematics solver based on the static levator scapulae anatomical base model to calculate the predicted length change curve of the levator scapulae between the attachment points of the cervical vertebrae and the scapula. The inverse kinematics solver is based on a static anatomical base model (including muscle fiber direction, attachment point location, etc.) and calculates the relationship between skeletal movement and muscle length changes in reverse to derive muscle deformation parameters (such as length and contraction speed) under specific movements. The predicted length change curve is a sequence of time-varying length changes of the levator scapulae muscle between its attachment points on the cervical spine and scapula, calculated by the inverse kinematics solver based on multi-angle dynamic optical motion capture data, and is used to quantify muscle deformation.

[0065] In this embodiment, the inverse kinematics solver uses a static anatomical base model as a reference and inputs multi-angle dynamic optical data (such as the time-varying sequence of the coordinates of the superior angle of the scapula). By minimizing the error between skeletal movement and muscle deformation, it inversely derives the length change of the levator scapulae muscle between its attachment points on the cervical vertebrae (C1-C4 transverse processes) and the scapula (medial border of the superior angle). For example, when the scapula is elevated, the distance between attachment points shortens. The solver iteratively adjusts the muscle path using optimization algorithms (such as Levenberg-Marquardt) to match the model's predicted length with the actual kinematic data, and finally outputs a curve showing the length changing over time (e.g., the length shortens from 10cm to 8.5cm within 0-2 seconds).

[0066] In step S132, the length change rate of the levator scapulae muscle is determined based on the predicted length change curve; Among them, the length change rate is the first derivative of the predicted length change curve, which represents the speed of length change of the muscle per unit time (e.g., cm / s), reflecting the dynamic load demand of muscle contraction.

[0067] In this embodiment of the disclosure, the first-order differential calculation is performed on the predicted length change curve to obtain the length change rate (i.e., the muscle contraction speed). For example, if the length curve shortens from 9.5cm to 8.8cm within 0.5-1.5 seconds, the change rate is (8.8-9.5) / (1.5-0.5) = -0.7cm / s, where the negative sign indicates shortening. The change rate reflects the dynamic load demand of the muscle: high-speed shortening (such as throwing movements) requires rapid neural activation, while low-speed shortening (such as maintaining posture) requires sustained low-intensity activation.

[0068] In step S133, the electrical signal sequence corresponding to the surface electromyography signal is input into a potential convolution model based on dual Hilbert transform to synthesize the theoretical high-density electromyography of the corresponding muscle on the skin surface. The dual Hilbert transform extracts the instantaneous phase and amplitude envelope of the signal through two Hilbert transforms, which is used to separate noise and effective components in the electromyography (EMG) signal, thereby improving the spatiotemporal resolution of high-density electromyography (HD-EMG). Theoretical HD-EMG is an electromyography distribution map with higher spatiotemporal resolution formed by recording the electrical activity signal of the muscle surface through a multi-channel electrode array (such as an 8×8 grid) and processing it (such as dual Hilbert transform), which can reflect the timing and intensity of nerve activation in different areas of the muscle.

[0069] In this embodiment, the dual Hilbert transform decomposes the frequency domain components of the surface electromyography (EMG) signal to extract the instantaneous phase (reflecting the timing of nerve discharges) and amplitude envelope (reflecting the number of recruited motor units). For example, the original signal is transformed by the first Hilbert transform to obtain an analytical signal, and then the amplitude of the analytical signal is transformed a second time to separate noise (high-frequency random fluctuations) and effective EMG components (low-frequency regular fluctuations), and finally synthesizes the theoretical HD-EMG of the skin surface (such as the spatiotemporal potential distribution map of an 8×8 electrode array).

[0070] In step S134, the activation sequence of the neural activation that drives the contraction of the levator scapulae muscle is estimated based on the theoretical high-density electromyography, the predicted length change curve, and the length change rate. In this embodiment, the onset / end time of neural activation is estimated by combining theoretical HD-EMG (reflecting neural discharge timing), predicted length change curves (reflecting muscle deformation requirements), and length change rate (reflecting dynamic load) using a convolutional neural network (CNN) or Bayesian inference model. For example, if high-frequency discharge (>50Hz) occurs in HD-EMG at 0.3 seconds, and the length change rate is -0.8cm / s (rapid shortening), the model determines the onset time of neural activation to be 0.3 seconds; if the discharge stops at 1.8 seconds, and the length change rate approaches 0, the end time is determined to be 1.8 seconds.

[0071] In step S135, the activation intensity of the target nerve that drives the contraction of the levator scapulae muscle is determined based on the length change rate.

[0072] In this embodiment of the disclosure, neural activation intensity is quantified based on the absolute value of the length change rate (reflecting the magnitude of muscle load) using a predefined activation intensity-load curve (e.g., a linear relationship: for every 10% increase in load, the activation intensity increases by 15%) or a machine learning model (e.g., a support vector machine). For example, if the length change rate is -0.7 cm / s (moderate load), the corresponding activation intensity is 60% (assuming the change rate is -1.2 cm / s when the maximum activation intensity is 100%), indicating that 60% of the motor units are recruited at this time.

[0073] The aforementioned technical solution achieves precise calculation of neuromuscular activation patterns through multimodal data fusion and biomechanical-electrophysiological coupling modeling. By combining optical motion capture (millimeter-level spatial accuracy) and surface electromyography (millisecond-level temporal accuracy), it can capture instantaneous dynamic changes in neural activation (such as brief activation during rapid movements). Through joint analysis of inverse kinematics and electromyographic signals, it links skeletal movement, muscle deformation, and neural control, avoiding the limitations of a single data source (such as electromyography being easily affected by fat thickness, and kinematics easily neglecting internal muscle deformation), thus improving the accuracy of intervention.

[0074] In a preferred embodiment, step S134, estimating the activation sequence of the neural activation driving the contraction of the levator scapulae muscle based on the theoretical high-density electromyography, the predicted length change curve, and the length change rate, includes: In step S1341, the predicted length values ​​of the levator scapulae muscle at different times are determined based on the predicted length change curve. In this embodiment, the levator scapulae length value for each time point (e.g., every 10 ms) is extracted from the predicted length change curve output by the inverse kinematics solver. For example, if the curve covers 0-2 seconds, 200 discrete length values ​​can be obtained (e.g., 10 cm at 0 seconds, 9.98 cm at 0.01 seconds), forming a time-length sequence. This sequence provides raw data on muscle deformation for subsequent analysis, and its accuracy depends on the sampling frequency of the optical motion capture (e.g., 1000 Hz) and the iterative error of the solver (typically <1 mm).

[0075] In step S1342, a model relating length to muscle contraction state is established by combining muscle physiological characteristics, and the degree of contraction of the levator scapulae muscle at each time point is analyzed to obtain a length feature sequence that reflects the dynamic changes in the contraction of the levator scapulae muscle. Among them, the length feature sequence combines muscle physiological characteristics (such as length-tension relationship) to convert the predicted length value into a sequence that reflects the degree of muscle contraction (such as shortening ratio, tension level), and is used to quantify the dynamic deformation state of the muscle.

[0076] In this embodiment of the disclosure, a mathematical relationship between length and contraction state is constructed based on muscle physiological characteristics (such as the Hill model). For example, assuming the optimal length of the levator scapulae muscle is 9.5 cm, when the length is >9.5 cm, it is in a stretched state (reduced tension), and when the length is <9.5 cm, it is in a shortened state (increased tension), and the shortening ratio is linearly related to the tension (e.g., a 10% shortening corresponds to a 20% increase in tension). Through this model, the predicted length value is converted into the degree of contraction (e.g., a shortening ratio of 0-100%), forming a length feature sequence.

[0077] In step S1343, the length change amplitude of the levator scapulae muscle per unit time is calculated based on the length change rate; In this embodiment of the disclosure, the predicted length change curve is differentially calculated to obtain the length difference between adjacent time points (e.g., ΔL = L(t + Δt) - L(t)), which is then divided by the time interval (Δt) to obtain the length change rate (e.g., ΔL / Δt = -0.2cm / 0.01s = -20cm / s). This value reflects the deformation rate of the muscle per unit time and is used to distinguish between static contraction (change rate ≈ 0) and dynamic contraction (e.g., the change rate can reach -50cm / s in a throwing motion).

[0078] In step S1344, the acceleration or deceleration phase of the levator scapulae contraction is determined based on the length change amplitude and the length feature sequence, thereby obtaining the length change rate information of the levator scapulae contraction. In this embodiment, the dynamic characteristics of muscle contraction are analyzed by combining the amplitude of length change and the length feature sequence. For example, if the length feature sequence shows that the muscle shortens from 10cm to 9cm (a 10% reduction), and the rate of change of length increases from -10cm / s to -30cm / s, it is determined to be in the accelerated contraction phase (enhanced neural activation); if the rate of change decreases from -30cm / s to -10cm / s, it is determined to be in the deceleration phase (weakened neural activation). The phase division is achieved through threshold segmentation (e.g., an absolute value of the rate of change > 15cm / s indicates acceleration) or derivative analysis (e.g., a second derivative of the rate of change > 0 indicates acceleration).

[0079] In step S1345, based on the dimensionally unified basic electromyographic features characterized by the theoretical high-density electromyography, the length feature sequence, and the length change rate information, the activation sequence of the neural activation driving the contraction of the levator scapulae muscle is estimated.

[0080] In this embodiment, the basic electromyographic characteristics of theoretical HD-EMG (such as root mean square value RMS and zero-crossing rate ZC) are used as electrophysiological indicators of neural activation. Combined with length feature sequences (reflecting muscle deformation requirements) and length change rate information (reflecting dynamic load), the activation timing is estimated through synergistic analysis. For example, if the HD-EMG RMS suddenly rises (> threshold 50 μV) in 0.5 seconds, and the length feature sequence shows that the muscle begins to shorten (from 10 cm → 9.8 cm) at this time, with a length change rate of -20 cm / s (accelerated contraction), then the neural activation initiation time is determined to be 0.5 seconds; if the RMS drops below the threshold in 1.2 seconds, and the muscle length stops shortening (change rate ≈ 0), then the end time is determined to be 1.2 seconds.

[0081] The aforementioned technical solution achieves accurate estimation of neural activation timing through multimodal data fusion and biomechanical-electrophysiological co-analysis. By combining muscle deformation (length change) with electrophysiological signals (HD-EMG), it avoids the limitations of single data sources (such as the susceptibility of electromyography to noise interference and the neglect of neural control in deformation analysis), making the activation timing more consistent with actual neural drive patterns. Through the information on the speed of length change (acceleration / deceleration phases), it can distinguish the differences in neural activation during static posture maintenance and dynamic movement execution (such as the different activation timings of slow arm raising and fast throwing), providing more detailed parameters for motor control research. It can provide physiologically reasonable control strategies for the diagnosis of movement disorders (such as delayed neural activation in patients with frozen shoulder), optimization of rehabilitation training (such as designing timing-matched electrical stimulation programs), and human-computer interaction (such as the synchronous control of neural signals in exoskeleton robots), thereby improving the effectiveness of interventions.

[0082] In a preferred embodiment, in step S14, the step of embedding the static levator scapulae anatomical base model into a pre-trained levator scapulae digital twin model according to the activation sequence and corresponding activation intensity of the target nerve to obtain a target levator scapulae digital twin includes: In step S141, the attention weights of the encoder in the pre-trained digital twin model of the levator scapulae muscle are fine-tuned according to the activation sequence and corresponding activation intensity of the target nerve. The attention weights are the weights of the target features when multiple parallel branches in the encoder fuse. Each of the multiple parallel branches performs feature extraction according to the input features to obtain the target features. Digital twins are virtual models built based on the physiological and mechanical properties of physical entities (such as the levator scapulae muscle). Driven by real-time data (such as neural activation signals), they simulate the dynamic behavior of the entity under specific conditions (such as deformation and stress distribution) to predict, optimize, or control the entity's state. Attention weights are parameters used in deep learning models (such as the Transformer encoder) to dynamically adjust the contribution of different input features to the output. Larger values ​​indicate a more important contribution of the feature to the current task, improving the model's ability to capture key information (such as the timing of neural activation). Latent space representation is the encoder's method of compressing high-dimensional input data (such as neural activation signals and muscle anatomy features) into low-dimensional vectors, retaining key features while removing redundant information, facilitating efficient generation of prediction results (such as deformation fields and stress distribution) by the decoder.

[0083] In this embodiment, the encoder consists of multiple parallel branches (such as convolutional layers and self-attention layers). Each branch independently extracts different dimensions of information (such as temporal patterns and spatial structures) from the input features (such as neural activation timing and muscle anatomy). Attention weights are generated using the Softmax function, ranging from [0,1], and the sum of all branch weights is 1. Based on the activation timing (e.g., activation begins at 0.2 seconds) and intensity (e.g., a weight of 0.8 corresponds to high-intensity activation), the weight values ​​of the corresponding branches are adjusted (e.g., the weight of the temporal branch increases from 0.3 to 0.6, and the weight of the anatomy branch decreases from 0.7 to 0.4), making the model focus more on features related to the current neural activation. The fusion layer sums the target features of each branch according to the weights to generate a latent space representation (e.g., a 128-dimensional vector). This representation retains the dynamic information of neural activation while incorporating the static features of muscle anatomy.

[0084] In the pre-trained digital twin model of the levator scapulae muscle, the decoder is connected to the encoder. The fusion layer of the encoder is used to fuse the target features of the multiple parallel branches according to the attention weights to form a latent space representation. The decoder is used to take the latent space representation as input and output the predicted three-dimensional deformation field data and the predicted muscle stress-strain distribution data of the levator scapulae muscle. The predicted 3D deformation field data describes the vector field of spatial position changes at various points in the muscle under neural activation, reflecting the geometric changes in muscle contraction / stretching (e.g., a point moves from coordinates (1,2,3) to (1.1,1.9,3.2)). The predicted stress-strain distribution data quantifies the mechanical state of various points within the muscle. Stress represents the force per unit area (e.g., 10 kPa), and strain represents the deformation ratio (e.g., 5% shortening). Together, they reflect the mechanical response characteristics of the muscle.

[0085] For example, the target nerve is the accessory nerve branch of the levator scapulae muscle, with an activation sequence starting at 0.3 seconds and lasting for 1 second, and a high activation intensity (corresponding to a weight of 0.7). In the encoder, the temporal branch extracts the start time (0.3 seconds) and duration (1 second) of the activation signal, while the anatomical branch extracts the initial muscle length (10 cm) and attachment point location (C1 cervical vertebrae to the superior angle of the scapula). By adjusting the attention weights (increasing the weight of the temporal branch from 0.4 to 0.7, and decreasing the weight of the anatomical branch from 0.6 to 0.3), the fusion layer generates a latent space representation, in which the temporal features have a higher proportion, making the decoder more focused on the impact of activation dynamics on deformation.

[0086] In step S142, the static levator scapulae anatomical base model is embedded into the downstream output of the decoder to obtain the target levator scapulae digital twin; In this embodiment, the decoder takes the latent space representation as input and generates predicted three-dimensional deformation field data (such as the displacement vector of each point on the muscle surface) and stress-strain distribution data (such as the stress value of each point inside the muscle) through fully connected layers or deconvolutional layers. The static levator scapulae anatomical base model includes the initial geometry of the muscle (such as length, width, and attachment point location) and tissue structure (such as muscle fiber arrangement direction and elastic modulus). The anatomical base model is embedded downstream of the decoder, i.e., driven by the predicted deformation field and stress-strain data, the geometry of the anatomical model is adjusted (such as shortening the length and changing the curvature) and tissue parameters (such as updating the elastic modulus to match the stress state) through finite element analysis or geometric transformation algorithms (such as Lagrange interpolation). Finally, the output is a muscle tissue morphology result that integrates dynamic mechanical response and static anatomical features (such as visualizing the morphology and internal stress distribution of the muscle after contraction).

[0087] The static levator scapulae anatomical base model is used to perform muscle geometry transformation and tissue construction based on the predicted three-dimensional deformation field data and the predicted muscle stress and strain distribution data, and outputs and displays the muscle tissue morphology results of the levator scapulae.

[0088] For example, the decoder outputs predicted deformation fields (e.g., mid-muscle displacement +2cm) and stress distribution (e.g., mid-muscle stress 15kPa). The static anatomical model is initially 10cm long, then shortened to 8cm after embedding based on the deformation field, and the muscle fiber elastic modulus is adjusted according to the stress distribution (e.g., mid-muscle modulus increases from 100MPa to 120MPa). Finally, it outputs the muscle morphology after contraction (8cm length, increased curvature) and visualization of internal stress (high-stress area in red in the middle).

[0089] The aforementioned technical solution achieves high-precision construction of a digital twin of the levator scapulae muscle through the deep fusion of dynamic neural signals and static anatomical models. Fine-tuning of the encoder's attention weights enables the model to dynamically capture the impact of the timing and intensity of neural activation on muscle deformation (e.g., rapid activation leading to more intense contraction). The deformation field and stress distribution output by the decoder more closely match actual physiological and mechanical responses, avoiding the limitations of a single anatomical model or static electromyography model. The static anatomical base model can be customized based on individual differences (e.g., muscle length, attachment point location), combined with dynamic neural signal drive, to generate a digital twin that conforms to individual characteristics. This provides a precise simulation tool for personalized rehabilitation training (e.g., optimizing muscle activation patterns for patients with frozen shoulder) or surgical planning (e.g., simulating deformation after muscle resection). The output muscle tissue morphology results (e.g., 3D deformation animation, stress heatmap) can intuitively demonstrate the dynamic relationship between neural activation and muscle deformation, helping doctors or patients understand muscle dysfunction (e.g., insufficient contraction due to delayed neural activation), improving the communication efficiency and acceptability of treatment plans.

[0090] In a preferred embodiment, step S11, acquiring static magnetic resonance imaging data of the target user's shoulder and neck region, multi-angle dynamic optical motion capture data, and surface electromyographic signals of the target muscle group, includes: In step S111, a magnetic resonance diffusion tensor imaging sequence is used to obtain the conventional anatomical structure of the shoulder and neck region and extract the orientation and diffusion anisotropy fraction information of the levator scapulae muscle fibers to obtain an initial medical image. Then, the initial medical image is subjected to semantic segmentation of tissue structure and three-dimensional mesh generation to obtain the static magnetic resonance image data. Among them, magnetic resonance diffusion tensor imaging (DTI) measures the diffusion anisotropy of water molecules in tissues (such as the diffusion speed along the direction of muscle fibers is much faster than in the vertical direction), quantifies the tissue microstructure (such as the orientation and integrity of muscle fibers), and outputs the diffusion anisotropy score (FA, ranging from 0 to 1, with higher values ​​indicating stronger anisotropy and more regular muscle fiber arrangement).

[0091] Semantic segmentation is a computer vision task that classifies each pixel / voxel in a medical image into a specific tissue type (such as muscle, bone, fat) and generates a labeled segmentation map.

[0092] In this embodiment, magnetic resonance diffusion tensor imaging (MRI) applies a gradient magnetic field pulse to induce a phase change in the diffusion of water molecules. The difference in diffusion velocity in different directions reflects the tissue microstructure. The muscle fibers of the levator scapulae muscle follow the main diffusion direction, and the FA value quantifies the regularity of the muscle fiber arrangement (e.g., FA=0.7 indicates a highly regular arrangement). The initial medical image contains T1-weighted (anatomical structure) and DTI (diffusion parameter) multimodal data. Semantic segmentation uses a deep learning model (e.g., U-Net) to identify muscles, bones, and other tissues, generating a segmentation mask. The 3D mesh generation uses the Marching Cubes algorithm to convert the segmentation mask into a 3D mesh model composed of vertices, edges, and faces, preserving the muscle geometry (e.g., length, curvature) and spatial position (e.g., attachment relationship with the cervical spine), ultimately outputting static magnetic resonance imaging data.

[0093] In step S112, a cluster of marker points is used to capture the three-dimensional motion trajectory of the cervical spine to thoracic spine, clavicle, and scapula when the target user performs the full range of motion of the levator scapulae muscle, thereby obtaining motion capture data. The motion capture data is then bound to a skeleton and a coordinate system to obtain the multi-angle dynamic optical motion capture data. The marker cluster consists of an array of multiple reflective markers (such as spherical patches with a diameter of 5 mm) that are attached to the surface of the target bone (such as the cervical spine or scapula). The markers are tracked synchronously by multiple cameras of an optical motion capture system (such as Vicon) to record the coordinate changes of the markers in three-dimensional space, reflecting the bone's movement trajectory.

[0094] In this embodiment, marker clusters are affixed to the surfaces of key bones such as the cervical vertebrae C1-C7, clavicle, and superior / inferior angles of the scapula. When the target user performs levator scapulae muscle movements (such as shoulder elevation and retraction), multiple infrared cameras (e.g., 8 cameras, 100Hz sampling rate) of the optical motion capture system simultaneously capture the reflected light from the markers and calculate the three-dimensional coordinates of each marker using triangulation principles. Skeletal binding maps the marker coordinates to a predefined skeletal model (e.g., 7 cervical vertebrae and 1 scapula), establishing kinematic relationships between bones (e.g., the rotation angle of the scapula relative to the thoracic ribcage). The coordinate system is based on the world coordinate system to eliminate differences in the perspectives of different cameras, ensuring spatial consistency of the motion trajectory, and finally outputting three-dimensional motion trajectory data of the cervical to thoracic vertebrae, clavicle, and scapula.

[0095] In step S113, an initial electromyographic signal is obtained by covering the upper trapezius muscle, the surface projection area of ​​the levator scapulae muscle, and the rhomboid muscle region with an electrode array. The electromyographic signal is then bandpass filtered, rectified, and decomposed into a motor unit firing sequence for the levator scapulae muscle using a blind source separation algorithm to obtain the surface electromyographic signal of the target muscle group.

[0096] Among them, blind source separation separates independent source signals (such as the firing sequence of a single muscle motor unit) from mixed signals (such as the superposition of electromyographic signals from multiple muscles) without prior knowledge of the source signal characteristics.

[0097] In this embodiment, an electrode array (e.g., an 8-channel flexible electrode) covers the upper trapezius muscle, the surface projection area of ​​the levator scapulae muscle (approximately C2 of the cervical spine to the superior angle of the scapula), and the rhomboid muscle region, acquiring bioelectrical signals (initial electromyographic signals, frequency range 5-500Hz) generated during muscle contraction. Bandpass filtering (e.g., 20-450Hz) removes low-frequency motion artifacts (e.g., skin friction) and high-frequency noise (e.g., power supply interference); rectification converts the AC signal into an absolute value signal, highlighting the amplitude characteristics of the electromyography; blind source separation (e.g., the FastICA algorithm), based on the assumption of signal statistical independence, separates the firing sequence of the motor units of the levator scapulae muscle (i.e., the time sequence of discharge of a single motor unit) from the mixed electromyography, reflecting the timing (e.g., discharge begins at 0.2 seconds) and intensity (e.g., discharge frequency 50Hz) of muscle activation, and finally outputs the surface electromyographic signal for the levator scapulae muscle.

[0098] The above-mentioned technical solution achieves comprehensive acquisition of physiological, motor, and electrophysiological information in the shoulder and neck region through multimodal data fusion. Static magnetic resonance imaging provides anatomical details of the levator scapulae muscle (such as muscle fiber orientation and attachment point location), dynamic motion capture data reflects skeletal movement trajectory (such as scapular upward rotation angle), and surface electromyography (EMG) signals quantify muscle activation intensity (such as discharge frequency). The combination of these three can reveal the dynamic interaction mechanism of the "neuromuscular-skeleton" system (such as EMG activation driving skeletal movement, and anatomical structure limiting the range of motion). Magnetic resonance imaging covers deep muscles (such as the levator scapulae muscle), motion capture captures superficial skeletal movement, and EMG signals reflect muscle electrical activity. Multi-level data complementarity avoids the limitations of a single modality (such as motion capture being unable to distinguish the source of muscle activation, and EMG signals lacking spatial localization). The acquired data can support the design of personalized rehabilitation programs (such as adjusting the levator scapulae muscle activation pattern for patients with frozen shoulder), surgical planning (such as simulating motor compensation after muscle resection), or sports injury assessment (such as analyzing the correlation between EMG abnormalities and movement trajectory deviations).

[0099] The present invention also provides a muscle tissue morphology display device based on a digital twin of the levator scapulae muscle, see [link to relevant documentation]. Figure 4 As shown, it includes: an acquisition module 410, configured to acquire static magnetic resonance imaging data of the target user's shoulder and neck region, multi-angle dynamic optical motion capture data, and surface electromyography signals of the target muscle group, and construct multimodal levator scapulae physiological data, wherein the target muscle group includes the levator scapulae and synergistic muscle groups that coordinate the movement of the levator scapulae; and a construction module 420, configured to construct a static levator scapulae anatomical base model based on the static magnetic resonance imaging data, using image segmentation and three-dimensional reconstruction technology, wherein the static levator scapulae anatomical base model is based on the geometric shape of the levator scapulae, the attachment point of the levator scapulae, the direction of the muscle fibers, the first spatial relationship between the levator scapulae and adjacent bones, and the second spatial relationship between the levator scapulae and adjacent ligaments. The system comprises: a calculation module 430 configured to calculate the activation sequence and corresponding activation intensity of the target nerve driving the contraction of the levator scapulae muscle under different movements based on the multi-angle dynamic optical motion capture data and the surface electromyography signal; an embedding module 440 configured to embed the static levator scapulae muscle anatomical base model into a pre-trained levator scapulae muscle digital twin model based on the activation sequence and corresponding activation intensity of the target nerve to obtain a target levator scapulae muscle digital twin; and a response output module 450 configured to, in response to the acquired input motion command, input the input motion command into the target levator scapulae muscle digital twin and output and display the muscle tissue morphology results of the levator scapulae muscle.

[0100] In a preferred embodiment, the construction module 420 is configured to: automatically segment the image sequence corresponding to the static magnetic resonance imaging data using a deep convolutional neural network, identify the muscle-tendon boundary of the levator scapulae muscle, the attachment points of the muscle to the transverse processes of the cervical vertebrae and the superior angle of the scapula, and obtain identification result data; based on the identification result data, reconstruct the spatial path and curvature of the main muscle fiber bundles within the muscle volume using a fiber tracing algorithm to obtain an initial muscle model, wherein the initial muscle model is a set of multiple representative muscle fiber bundles refined from a uniform entity; based on the initial muscle model, generate a non-uniform tetrahedral mesh three-dimensional model, and assign initial values ​​of anisotropic hyperelastic material properties to each unit of each mesh in the tetrahedral mesh three-dimensional model according to the anatomical location to which each unit belongs and the local diffuse anisotropy fraction along the muscle fiber direction; based on the initial property values, add a muscle fiber direction vector field to each unit of each mesh in the tetrahedral mesh three-dimensional model to obtain the static levator scapulae muscle anatomical base model.

[0101] In a preferred embodiment, the construction module 420 is configured to: determine the anatomical location of each unit by comparing the geometric position, shape features, and relative relationship with surrounding structures of each unit in the tetrahedral mesh 3D model using a pre-built anatomical location database; determine the muscle fiber direction of the anatomical location of each unit based on the reference physiological structure and reference motor function of the levator scapulae muscle; calculate a first diffusion coefficient along the muscle fiber direction and a second diffusion coefficient perpendicular to the muscle fiber direction for each unit based on the muscle fiber direction of the anatomical location of each unit and diffusion tensor imaging data; determine the degree of diffusion difference of the unit in different directions based on the first diffusion coefficient and the second diffusion coefficient corresponding to each unit, and obtain a local diffusion anisotropy score; and assign corresponding initial attribute values ​​to each unit based on the anatomical location of each unit and the local diffusion anisotropy score, with reference to a mechanical property database of anisotropic hyperelastic materials.

[0102] In a preferred embodiment, the construction module 420 is configured to: acquire initial attribute values ​​for each unit in the tetrahedral mesh 3D model, analyze their correlation with muscle fiber characteristics, and determine the distribution information of anisotropy and elastic modulus in different directions; based on the distribution information of each anisotropy and elastic modulus in different directions, formulate construction rules for a muscle fiber direction vector field, wherein the construction rules specify the vector magnitudes in the muscle fiber direction vector field, the association methods of each anisotropy fraction, and the muscle fiber physiological orientation information followed by the vector directions; and, based on the construction rules, for the tetrahedral mesh... For each unit in the 3D mesh model, a corresponding muscle fiber direction vector field is added with the corresponding initial attribute value as the target. Based on the muscle fiber direction vector field, the tetrahedral mesh 3D model is compared with the actual muscle tissue structure features. If the comparison result does not meet the preset conditions, the initial attribute value is readjusted, and the steps of obtaining the initial attribute value of each unit in the 3D tetrahedral mesh model and comparing the tetrahedral mesh 3D model with the actual muscle tissue structure features are repeated until the comparison result meets the preset conditions, thus obtaining the static levator scapulae anatomical base model.

[0103] In a preferred embodiment, the solution module 430 is configured to: input the multi-angle dynamic optical motion capture data into an inverse kinematics solver based on the static levator scapulae anatomical base model to calculate the predicted length change curve of the levator scapulae between the attachment points of the cervical vertebrae and the scapula; determine the length change rate of the levator scapulae based on the predicted length change curve; input the electrical signal sequence corresponding to the surface electromyography signal into a potential convolution model based on double Hilbert transform to synthesize a theoretical high-density electromyography of the corresponding muscle on the skin surface; estimate the activation sequence of the nerve activation driving the contraction of the levator scapulae based on the theoretical high-density electromyography, the predicted length change curve, and the length change rate; and determine the activation intensity of the target nerve driving the contraction of the levator scapulae based on the length change rate.

[0104] In a preferred embodiment, the calculation module 430 is configured to: determine the predicted length value of the levator scapulae muscle at different times based on the predicted length change curve; establish a relationship model between length and muscle contraction state by combining muscle physiological characteristics, analyze the degree of contraction of the levator scapulae muscle at each time, and obtain a length feature sequence reflecting the dynamic changes in the contraction of the levator scapulae muscle; calculate the length change amplitude of the levator scapulae muscle per unit time based on the length change rate; determine the acceleration or deceleration phase of the levator scapulae muscle contraction based on the length change amplitude and the length feature sequence, and obtain the length change rate information of the levator scapulae muscle contraction; and estimate the activation sequence of the neural activation driving the contraction of the levator scapulae muscle based on the dimensionally unified basic electromyographic characteristics represented by the theoretical high-density electromyography, the length feature sequence, and the length change rate information.

[0105] In a preferred embodiment, the embedding module 440 is configured to: fine-tune the attention weights of the encoder in the pre-trained levator scapulae digital twin model according to the activation sequence and corresponding activation intensity of the target nerve, wherein the attention weights are the weights for fusing target features of multiple parallel branches in the encoder, each of the multiple parallel branches performing feature extraction based on input features to obtain the target features; wherein the decoder in the pre-trained levator scapulae digital twin model is connected to the encoder, and the fusion layer of the encoder is used to fuse the target features of the multiple parallel branches according to the attention weights to form a latent space representation; the decoder is used to output the predicted three-dimensional deformation field data and predicted muscle stress-strain distribution data of the levator scapulae as input; embedding the static levator scapulae anatomical base model downstream of the decoder output to obtain the target levator scapulae digital twin; wherein the static levator scapulae anatomical base model is used to perform muscle geometry transformation and tissue construction based on the predicted three-dimensional deformation field data and the predicted muscle stress-strain distribution data, and output and display the muscle tissue morphology results of the levator scapulae.

[0106] In a preferred embodiment, the acquisition module 410 is configured to: acquire the conventional anatomical structure of the shoulder and neck region and extract the orientation and diffusion anisotropy fraction information of the levator scapulae muscle fibers using a magnetic resonance diffusion tensor imaging sequence to obtain an initial medical image; perform semantic segmentation of the tissue structure and three-dimensional mesh generation on the initial medical image to obtain the static magnetic resonance image data; capture the three-dimensional motion trajectory of the cervical spine to thoracic spine, clavicle, and scapula when the target user performs the full range of movements of the levator scapulae muscle using a marker cluster to obtain motion capture data; perform skeleton binding and coordinate system one on the motion capture data to obtain the multi-angle dynamic optical motion capture data; acquire initial electromyographic signals by covering the upper trapezius muscle, the surface projection area of ​​the levator scapulae muscle, and the rhomboid muscle region with an electrode array; perform bandpass filtering and rectification on the electromyographic signals and decompose the motor unit firing sequence for the levator scapulae muscle using a blind source separation algorithm to obtain the surface electromyographic signals of the target muscle group.

[0107] This invention provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of any of the methods described in the foregoing embodiments.

[0108] This invention provides a muscle tissue morphology display device based on a digital twin of the levator scapulae muscle, comprising: a memory storing a computer program thereon; and a processor for executing the computer program in the memory to implement the steps of any of the methods described in the foregoing embodiments.

[0109] Figure 5 The muscle tissue morphology display device 100 based on a digital twin of the levator scapulae muscle, as shown, includes a processor 1001 and a memory 1003. The processor 1001 and the memory 1003 are connected, for example, via a bus 1002. Optionally, the muscle tissue morphology display device 100 based on a digital twin of the levator scapulae muscle may further include a communication component 1004, which can be used for data interaction between the device 100 and other devices, such as data transmission and / or data reception. It should be noted that in actual operation, the communication component 1004 is not limited to one, and the structure of this muscle tissue morphology display device 100 based on a digital twin of the levator scapulae muscle does not constitute a limitation on the embodiments of this application.

[0110] Processor 1001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 1001 may also be a combination that implements computing functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0111] Bus 1002 may include a pathway for transmitting information between the aforementioned components. Bus 1002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. Bus 1002 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 5 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0112] The memory 1003 may be ROM (Read Only Memory) or other types of static storage devices capable of storing static information and instructions, RAM (Random Access Memory) or other types of dynamic storage devices capable of storing information and instructions, or EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media, other magnetic storage devices, or any other medium capable of carrying or storing program code and capable of being read by a computer, without limitation herein.

[0113] The memory 1003 is used to store program code for executing embodiments of the present disclosure, and its execution is controlled by the processor 1001. The processor 1001 is used to execute the program code stored in the memory 1003 to implement the steps shown in the aforementioned embodiments of the muscle tissue morphology display method based on the levator scapulae digital twin.

[0114] The preferred embodiments of the present disclosure have been described in detail above with reference to the accompanying drawings. However, the present disclosure is not limited to the specific details of the above embodiments. Within the scope of the technical concept of the present disclosure, various changes, modifications, substitutions and variations can be made to these embodiments, and all such changes, modifications, substitutions and variations fall within the protection scope of the present disclosure.

[0115] It should also be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction, and such combinations should also be considered as part of this disclosure. To avoid unnecessary repetition, this disclosure will not further describe the various possible combinations. The technical scope of this application is not limited to the contents of the specification, but must be determined according to the scope of the claims.

Claims

1. A method for displaying muscle tissue morphology based on a digital twin of the levator scapulae muscle, characterized in that, The method includes: The system acquires static magnetic resonance imaging data of the target user's shoulder and neck region, multi-angle dynamic optical motion capture data, and surface electromyography signals of the target muscle group to construct multimodal physiological data of the levator scapulae muscle. The target muscle group includes the levator scapulae muscle and synergistic muscle groups that work in conjunction with the movement of the levator scapulae muscle. Based on the static magnetic resonance imaging data, a static levator scapulae anatomical base model is constructed using image segmentation and three-dimensional reconstruction techniques. The static levator scapulae anatomical base model is constructed based on the geometric shape of the levator scapulae, the attachment point of the levator scapulae, the direction of the muscle fibers, the first spatial relationship between the levator scapulae and adjacent bones, and the second spatial relationship between the levator scapulae and adjacent ligaments. Based on the multi-angle dynamic optical motion capture data and the surface electromyography signal, the activation sequence and corresponding activation intensity of the target nerve driving the contraction of the levator scapulae muscle under different movements are calculated. Based on the activation sequence and corresponding activation intensity of the target nerve, the static levator scapulae anatomical base model is embedded into the pre-trained levator scapulae digital twin model to obtain the target levator scapulae digital twin. In response to the acquired input motion command, the input motion command is input into the target levator scapulae digital twin, and the muscle tissue morphology result of the levator scapulae is output and displayed.

2. The method according to claim 1, characterized in that, The process of constructing a static levator scapulae anatomical base model based on the static magnetic resonance imaging data, using image segmentation and 3D reconstruction techniques, includes: The image sequence corresponding to the static magnetic resonance imaging data is automatically segmented using a deep convolutional neural network to identify the muscle-tendon boundary of the levator scapulae muscle, the attachment point of the muscle to the transverse process of the cervical vertebrae and the superior angle of the scapula, and to obtain the identification result data. Based on the identification results, a fiber tracing algorithm is used to reconstruct the spatial paths and curvatures of the main muscle fiber bundles within the muscle volume to obtain an initial muscle model. The initial muscle model is a set of multiple representative muscle fiber bundles that refine a uniform entity. Based on the initial muscle model, a non-uniform tetrahedral mesh three-dimensional model is generated. Based on the anatomical location to which each unit of each mesh in the tetrahedral mesh three-dimensional model belongs and the local diffuse anisotropy fraction along the muscle fiber direction, an anisotropic initial value of hyperelastic material properties is assigned to each unit of each mesh in the tetrahedral mesh three-dimensional model. Based on the initial values ​​of the attributes, a muscle fiber direction vector field is added to each cell of each mesh in the tetrahedral mesh 3D model to obtain the static levator scapulae anatomical base model.

3. The method according to claim 2, characterized in that, The process of assigning anisotropic initial values ​​for the properties of the hyperelastic material to each element of each grid in the tetrahedral mesh 3D model, based on the anatomical location to which each element belongs and the local diffuse anisotropy fraction along the muscle fiber direction, includes: Based on a pre-built database of anatomical locations, the geometric position, shape features, and relative relationship with surrounding structures of each cell in the tetrahedral mesh 3D model are compared to determine the anatomical location to which each cell belongs. Based on the reference physiological structure and reference motor function of the levator scapulae muscle, the direction of muscle fibers in the anatomical location to which each unit belongs is determined; Based on the muscle fiber direction of the anatomical location to which each unit belongs, and combined with diffusion tensor imaging data, calculate the first diffusion coefficient of each unit along the muscle fiber direction and the second diffusion coefficient in the direction perpendicular to the muscle fiber direction. Based on the first diffusion coefficient and the second diffusion coefficient corresponding to each unit, the degree of diffusion difference of the unit in different directions is determined, and the local diffusion anisotropy score is obtained. Based on the anatomical location to which each unit belongs and the local diffusion anisotropy fraction, and referring to the mechanical property database of anisotropic hyperelastic materials, each unit is assigned a corresponding initial property value.

4. The method according to claim 2, characterized in that, The step of adding a muscle fiber direction vector field to each cell of each mesh in the tetrahedral mesh 3D model according to the initial value of the attribute to obtain the static levator scapulae anatomical base model includes: The initial values ​​of the properties of each element in the tetrahedral mesh 3D model are obtained, and their correlation with muscle fiber properties is analyzed to determine the degree of anisotropy and the distribution information of elastic modulus in different directions. Based on the distribution information of each degree of anisotropy and elastic modulus in different directions, a construction rule for the muscle fiber direction vector field is formulated. The construction rule is used to specify the vector magnitude in the muscle fiber direction vector field, the association method of each anisotropy fraction, and the muscle fiber physiological orientation information followed by the direction of the vector. According to the construction rules, for each unit in the tetrahedral mesh 3D model, with the corresponding initial attribute value as the target, a corresponding muscle fiber direction vector field is added; Based on the muscle fiber direction vector field, the tetrahedral mesh 3D model is compared with the actual muscle tissue structure features; If the comparison result does not meet the preset conditions, the initial value of the attribute is readjusted, and the steps of obtaining the initial value of the attribute of each unit in the tetrahedral mesh three-dimensional model and comparing the tetrahedral mesh three-dimensional model with the actual muscle tissue structure features are repeated until the comparison result meets the preset conditions, and the static levator scapulae anatomical base model is obtained.

5. The method according to claim 1, characterized in that, The step of calculating the activation sequence and corresponding activation intensity of the target nerve driving the levator scapulae muscle contraction under different movements based on the multi-angle dynamic optical motion capture data and the surface electromyography signals includes: The multi-angle dynamic optical motion capture data is input into the inverse kinematics solver based on the static levator scapulae anatomical base model to calculate the predicted length change curve of the levator scapulae between the attachment points of the cervical vertebrae and the scapula. The rate of change in length of the levator scapulae muscle is determined based on the predicted length change curve. The electrical signal sequence corresponding to the surface electromyography signal is input into a potential convolution model based on dual Hilbert transform to synthesize the theoretical high-density electromyography of the corresponding muscle on the skin surface. Based on the theoretical high-density electromyography, the predicted length change curve, and the length change rate, the activation sequence of the neural activation that drives the contraction of the levator scapulae muscle is estimated. The activation intensity of the target nerve driving the contraction of the levator scapulae muscle is determined based on the rate of change in length.

6. The method according to claim 5, characterized in that, The step of estimating the activation sequence of the neural activation driving the contraction of the levator scapulae muscle based on the theoretical high-density electromyography, the predicted length change curve, and the length change rate includes: Based on the predicted length change curve, the predicted length values ​​of the levator scapulae muscle at different times are determined; A model relating length to muscle contraction state was established by combining muscle physiological characteristics. The degree of contraction of the levator scapulae muscle at each time point was analyzed to obtain a length feature sequence that reflects the dynamic changes in the contraction of the levator scapulae muscle. Based on the stated rate of change in length, the amplitude of the length change of the levator scapulae muscle per unit time is calculated; Based on the length change amplitude and the length feature sequence, determine the acceleration or deceleration phase of the levator scapulae contraction, and obtain the length change rate information of the levator scapulae contraction. Based on the dimensionally unified basic electromyographic features characterized by the theoretical high-density electromyography, the length feature sequence, and the length change rate information, the activation sequence of the neural activation driving the contraction of the levator scapulae muscle is estimated.

7. The method according to claim 1, characterized in that, The process involves embedding the static levator scapulae anatomical base model into a pre-trained levator scapulae digital twin model based on the activation sequence and corresponding activation intensity of the target nerve, to obtain a target levator scapulae digital twin, including: Based on the activation sequence and corresponding activation intensity of the target nerve, the attention weights of the encoder in the pre-trained digital twin model of the levator scapulae muscle are fine-tuned. The attention weights are the weights of the target features when multiple parallel branches in the encoder fuse. Each of the multiple parallel branches performs feature extraction based on the input features to obtain the target features. In the pre-trained digital twin model of the levator scapulae muscle, the decoder is connected to the encoder. The fusion layer of the encoder is used to fuse the target features of the multiple parallel branches according to the attention weights to form a latent space representation. The decoder is used to take the latent space representation as input and output the predicted three-dimensional deformation field data and the predicted muscle stress-strain distribution data of the levator scapulae muscle. The static levator scapulae anatomical base model is embedded into the downstream output of the decoder to obtain the digital twin of the target levator scapulae muscle; The static levator scapulae anatomical base model is used to perform muscle geometry transformation and tissue construction based on the predicted three-dimensional deformation field data and the predicted muscle stress and strain distribution data, and outputs and displays the muscle tissue morphology results of the levator scapulae.

8. The method according to any one of claims 1-7, characterized in that, The acquisition of static magnetic resonance imaging data of the target user's shoulder and neck region, multi-angle dynamic optical motion capture data, and surface electromyographic signals of the target muscle group includes: Using magnetic resonance diffusion tensor imaging sequence, the conventional anatomical structure of the shoulder and neck region is obtained, and the orientation and diffusion anisotropy fraction information of the levator scapulae muscle fibers are extracted to obtain the initial medical image. Then, the initial medical image is subjected to semantic segmentation of tissue structure and three-dimensional mesh generation to obtain the static magnetic resonance image data. By using a cluster of marker points, the three-dimensional motion trajectory of the cervical spine to thoracic spine, clavicle, and scapula is captured when the target user performs the full range of motion of the levator scapulae muscle, thus obtaining motion capture data. The motion capture data is then bound to a skeleton and a coordinate system to obtain the multi-angle dynamic optical motion capture data. Initial electromyographic (EMG) signals were acquired by covering the upper trapezius muscle, the surface projection area of ​​the levator scapulae muscle, and the rhomboid muscle region with an electrode array. The EMG signals were then bandpass filtered, rectified, and decomposed into the motor unit firing sequence for the levator scapulae muscle using a blind source separation algorithm, thereby obtaining the surface EMG signals of the target muscle group.

9. A muscle tissue morphology display device based on a digital twin of the levator scapulae muscle, characterized in that, include: The acquisition module is configured to acquire static magnetic resonance imaging data of the target user's shoulder and neck region, multi-angle dynamic optical motion capture data, and surface electromyography signals of the target muscle group, and construct multimodal physiological data of the levator scapulae muscle, wherein the target muscle group includes the levator scapulae muscle and synergistic muscle groups that coordinate with the movement of the levator scapulae muscle. The construction module is configured to construct a static levator scapulae anatomical base model based on the static magnetic resonance imaging data, using image segmentation and three-dimensional reconstruction technology. The static levator scapulae anatomical base model is constructed based on the geometric shape of the levator scapulae, the attachment point of the levator scapulae, the direction of the muscle fibers, the first spatial relationship between the levator scapulae and adjacent bones, and the second spatial relationship between the levator scapulae and adjacent ligaments. The calculation module is configured to calculate the activation sequence and corresponding activation intensity of the target nerve that drives the levator scapulae muscle to contract under different movements, based on the multi-angle dynamic optical motion capture data and the surface electromyography signal. The embedding module is configured to embed the static levator scapulae anatomical base model into a pre-trained levator scapulae digital twin model according to the activation sequence and corresponding activation intensity of the target nerve, thereby obtaining a target levator scapulae digital twin. The response output module is configured to respond to the acquired input motion command, input the input motion command into the target levator scapulae digital twin, and output and display the muscle tissue morphology result of the levator scapulae.

10. An electronic device, characterized in that, include: A memory on which computer programs are stored; A processor for executing the computer program in the memory to implement the steps of the method according to any one of claims 1-8.

Images