A rehabilitation management method for children with duchenne muscular dystrophy based on three-dimensional gait analysis combined with nsaa

Abstract

This invention relates to the field of rehabilitation treatment technology for pediatric neuromuscular diseases, specifically to a rehabilitation management method for children with DMD based on three-dimensional gait analysis combined with NSAA, comprising the following steps: S100, acquiring basic multimodal feature data: acquiring multidimensional temporal gait feature data of children with DMD and standardized coding vectors based on the Polaris Mobility Scale assessment; S200, based on a preset feature mapping matrix, performing consistency deviation analysis on the standardized coding vectors and the multidimensional temporal gait feature data. This invention constructs a closed-loop feedback control mechanism based on the gradient of temporal feature changes, which can adaptively trigger iterative updates of system hardware instructions according to the staged improvement or progressive compensatory needs of the child's function, thereby providing children with DMD with a closed-loop rehabilitation management system from objective data acquisition and accurate parameter calculation to dynamic control of external equipment.

Description

Technical Field

[0001] This invention relates to the field of rehabilitation treatment technology for pediatric neuromuscular diseases, specifically to a rehabilitation management method for children with DMD based on three-dimensional gait analysis combined with NSAA. Background Technology

[0002] Duchenne muscular dystrophy (DMD) is an X-linked recessive inherited neuromuscular disease that primarily affects male children. Its core pathological feature is progressive degeneration and necrosis of muscle fibers, leading to progressive muscle weakness and gait abnormalities from early childhood. This gradually progresses to joint contractures, loss of motor function, and ultimately affects respiratory and cardiac function, severely impacting the child's quality of life and lifespan. Rehabilitation management is a core component of comprehensive treatment for children with DMD, with the core goals of slowing muscle weakness decline, maintaining joint range of motion, protecting motor and respiratory function, and prolonging independent walking time.

[0003] However, existing rehabilitation management techniques for children with DMD have the following significant technical shortcomings:

[0004] (1) Subjectivity of assessment methods: The current mainstream clinical assessment tools used include the North Star Movement Scale (NSAA) and 6-minute walking distance. Among them, the NSAA relies on the assessor's subjective observation and scoring of the child's 17 motor functions, which is easily affected by the assessor's experience, operational standardization, and environmental factors. The objectivity and repeatability of the assessment results are insufficient, making it difficult to accurately reflect the child's true functional status.

[0005] (2) Gait data is missing: The existing assessment does not systematically collect spatiotemporal parameters (step length, step frequency, step speed), joint angles and torques and other kinematic / dynamic data during the child's walking process. It is impossible to accurately locate the specific muscle group distribution of muscle weakness and the quantitative degree of joint movement restriction, resulting in a lack of targeted basis for rehabilitation intervention.

[0006] (3) The programs are highly homogenized: most rehabilitation programs are based on general guidelines and do not take into account the individual gait characteristics, muscle strength distribution and functional grade differences of the children. They cannot achieve individualized adaptation of "one policy for one person" and the intervention effect is limited.

[0007] (4) Difficulty in quantifying efficacy: Due to the lack of integration analysis of objective data and subjective scores, rehabilitation efficacy can only be judged by changes in subjective scores or observation of clinical symptoms. It is difficult to quantify and evaluate the specific improvement effect of intervention measures on gait function and muscle strength, and cannot support the scientific and dynamic adjustment of rehabilitation strategies.

[0008] Therefore, there is an urgent need in this field for a rehabilitation management technology for children with DMD that can integrate objective gait data with subjective functional scores, achieve accurate assessment, personalized treatment plan development, and quantitative efficacy tracking, in order to fill the gap in existing technologies. Summary of the Invention

[0009] To achieve the above objectives, the present invention provides the following technical solution:

[0010] This invention provides a rehabilitation management method for children with DMD based on three-dimensional gait analysis combined with NSAA, comprising the following steps:

[0011] S100, Obtain basic multimodal feature data: Obtain multidimensional temporal gait feature data of children with DMD and standardized coding vectors based on the Polaris Movement Scale assessment. The multidimensional temporal gait feature data includes spatiotemporal parameters, joint angle data and joint torque data in natural walking state.

[0012] S200, Feature fusion evaluation based on dynamic weighting: Based on the preset feature mapping matrix, the standardized encoding vector and the multidimensional temporal gait feature data are subjected to consistency deviation analysis, and the dynamic weighting mechanism is triggered to perform feature fusion according to the deviation analysis results, and the target muscle group reduction index and joint activity limitation quantification value are output.

[0013] S300, Generate digital intervention and control parameters: Based on the target muscle group decline index and the joint mobility limitation quantification value, calculate and generate digital intervention parameters for external assistive devices, including the dynamic assist parameters of the rehabilitation robot and the structural adjustment parameters of the orthosis;

[0014] S400, Adaptive parameter update based on temporal gradient: Establish a longitudinal database based on the historical assessment sequence of the child, obtain multimodal feature data of the retest after a preset period to calculate the feature change gradient, and adaptively iterate and update the digital intervention parameters based on the feature change gradient.

[0015] Furthermore, the acquisition of multidimensional temporal gait feature data specifically includes:

[0016] Acquire continuous gait cycle data collected by a 3D spatial motion capture device, and parse the continuous gait cycle data into the following structured temporal features:

[0017] The spatiotemporal parameter set includes the stride length, stride frequency, and stride speed characteristic sequences of the child when walking naturally in the preset test space;

[0018] Three-dimensional joint kinematic sequence, including dynamic range curves of the hip, knee and ankle joints of the child in the sagittal, coronal and horizontal planes;

[0019] And joint dynamic torques, including calculated hip flexion / extension torques, knee flexion / extension torques, and ankle plantar flexion / dorsiflexion torques for the child.

[0020] Furthermore, obtaining the standardized coding vector based on the Polaris Mobility Scale assessment specifically includes:

[0021] The discrete score data of the child when performing 17 preset independent motor function tasks were obtained, and the discrete score data was marked with preset level labels according to the degree of functional completion.

[0022] The discrete scoring data is processed by one-hot encoding or dimensionality reduction mapping to convert the grade labels into machine-readable standardized encoding vectors with fixed dimensions to characterize the child's subjective motor function status.

[0023] Furthermore, prior to the feature fusion evaluation based on dynamic weighting, a data alignment preprocessing step for physiological scale differences is included, specifically:

[0024] Extract the child’s basic physiological parameters, which include at least age, height and weight;

[0025] Construct a scale compensation factor based on the aforementioned basic physiological parameters;

[0026] The spatiotemporal parameter set and the joint dynamic torque are normalized and scaled using the scale compensation factor to eliminate the dimensional shift caused by the physical scale differences between different children on the gait feature data, and to obtain aligned standard multidimensional temporal gait feature data.

[0027] Furthermore, the step of performing consistency deviation analysis on the standardized encoding vector and the multidimensional temporal gait feature data, and triggering a dynamic weighting mechanism for feature fusion based on the deviation analysis results, specifically includes:

[0028] By projecting the standardized encoding vector onto the same feature space as the standard multidimensional temporal gait feature data using the preset feature mapping matrix, an aligned subjective evaluation feature vector is obtained. With objective gait feature vector ;

[0029] Calculate the consistency deviation between the subjective evaluation feature vector and the objective gait feature vector. :

[0030]

[0031] in, The dimension of the feature space. and The objective gait feature vector and the subjective evaluation feature vector are respectively in the th... Dimensional components;

[0032] Based on the consistency deviation value Trigger the dynamic weighting mechanism to calculate dynamic objective weights for objective gait data. :

[0033]

[0034] in, The preset basic objective weight is set to 0.6; This represents the upper limit of the objective weight, with a value of 0.7. This is the preset deviation compensation coefficient;

[0035] Calculate the corresponding dynamic subjective weights This ensures that the value of the dynamic subjective weight is between 0.3 and 0.4.

[0036] Based on the calculated dynamic objective weights and dynamic subjective weights, the two sets of feature vectors are weighted and fused to calculate the fused feature vector. :

[0037] .

[0038] Furthermore, the output target muscle group reduction index and joint mobility limitation quantification value specifically include:

[0039] The fused feature vector is input into a preset feature decoupling model to separate and extract the muscle weakness feature subset and the joint limitation feature subset;

[0040] Gradient threshold determination is performed on the subset of muscle weakness features, and the target muscle group decline index is quantitatively output. The target muscle group decline index is used to discretize and characterize whether the hip extensor muscle group, knee extensor muscle group or ankle plantar flexor muscle group of the child is in a mild, moderate or severe decline state.

[0041] The kinematic space inverse solution is performed on the restricted feature subset of the joint to extract the motion limit boundary of the restricted joint on the corresponding anatomical plane, and the joint motion restriction quantization value is generated. The joint motion restriction quantization value includes the specific coordinates of the restricted part and the corresponding restricted angle.

[0042] Furthermore, the generation of digital intervention parameters for the external assistive device, when the external assistive device is a smart ankle-foot orthosis, specifically includes:

[0043] The joint range of motion limitation was analyzed to extract the child's current maximum ankle dorsiflexion angle during the gait cycle. ;

[0044] Based on a pre-set target dorsiflexion angle for healthy children of the same age With the current maximum ankle dorsiflexion angle Calculate the compensation amount for angular loss;

[0045] The dynamic dorsiflexion assist stiffness of the intelligent ankle-foot orthosis is calculated using the following formula. :

[0046]

[0047] in, The physical reference stiffness of the orthotics. The set stiffness compensation coefficient, The ankle plantar flexor muscle group reduction index is the corresponding index in the target muscle group reduction index.

[0048] The calculated dynamic backbend auxiliary stiffness The corresponding support phase angle range is converted and generated as the structural adjustment parameters of the intelligent ankle-foot orthosis.

[0049] Furthermore, the generation of digital intervention parameters for the external assistive device, when the external assistive device is a lower limb rehabilitation exoskeleton robot, specifically includes:

[0050] Based on the target muscle group decline index, the dynamic functional deficit level of the child's hip extensor muscles and knee extensor muscles is determined.

[0051] The child's gait cycle is divided into different auxiliary intervals, and the real-time auxiliary torque of the lower limb rehabilitation exoskeleton robot at the target joint is calculated using the following dynamic auxiliary equation. :

[0052]

[0053] in, The time phase of the gait cycle, To provide the reference joint torque for the standard gait in the corresponding time phase, The corresponding muscle group reduction index is the index of the joint. It is a non-linear exponential gain factor. This is the preset joint damping coefficient. The real-time joint angular velocity fed back by the sensors of the rehabilitation robot;

[0054] The real-time auxiliary torque The timing curves and associated motor trigger thresholds are compiled into the low-level control messages of the lower limb rehabilitation exoskeleton robot, which are used as the dynamic auxiliary parameters.

[0055] The digital intervention parameters also include respiratory impedance parameters for intelligent respiratory training devices, and the specific generation steps include:

[0056] Extract a subset of subjective features related to core trunk control capabilities from the standardized coding vector;

[0057] The subjective feature subset is weighted and calculated to evaluate and generate a thoracic expansion restriction coefficient;

[0058] Based on the chest expansion restriction coefficient, the target value of positive end-expiratory pressure (PEEP) and the opening pressure threshold of the inspiratory valve of the intelligent breathing training device are calculated, and the target positive pressure value and the opening pressure threshold are packaged and sent to the intelligent breathing training device.

[0059] Furthermore, the step of acquiring retested multimodal feature data after a preset period to calculate the feature change gradient specifically includes:

[0060] Extract the baseline fusion feature vector of the child at the initial time phase or the previous assessment cycle from the longitudinal database;

[0061] The retested multimodal feature data of the current time phase after intervention for at least one preset period is obtained, and the current fused feature vector is generated through the same processing flow as steps S100 to S200.

[0062] By comparing the current fused feature vector with the benchmark fused feature vector, the absolute changes in spatiotemporal parameters and the relative rate of change of the quantized value of joint activity limitation are extracted respectively, and they are spliced ​​together to form a change gradient vector containing multidimensional improvement trend or decline trend features.

[0063] Furthermore, the adaptive iterative update of the digital intervention parameters based on the feature change gradient specifically includes parsing the change gradient vector and executing the following closed-loop update strategy:

[0064] When it is determined that the joint activity improvement gradient in the changing gradient vector reaches a preset positive gain threshold, a degradation assistance command is generated. The degradation assistance command is used to reduce the current dynamic assistance parameters of the lower limb rehabilitation exoskeleton robot by a preset step size, or to reduce the preset daytime working time cycle of the intelligent ankle-foot orthosis.

[0065] When the target muscle group's decline gradient in the changing gradient vector exceeds a preset decline risk threshold, a compensatory enhancement mechanism is triggered, automatically generating an upgrade assistance command. The upgrade assistance command is used to increase the dynamic dorsiflexion assist stiffness of the intelligent ankle-foot orthosis or increase the target damping coefficient of the lower limb rehabilitation exoskeleton robot.

[0066] Beneficial effects

[0067] This invention establishes a multimodal data fusion model encompassing three-dimensional temporal gait features and NSAA standardized coding, and introduces a dynamic weighting algorithm based on consistency bias. This effectively overcomes the bias risks inherent in traditional single subjective scale assessments, significantly improving the objectivity and accuracy of quantifying the motor function status of children with DMD. Simultaneously, it utilizes physiological scale compensation factors to eliminate dimensional shifts caused by individual developmental differences among children, precisely decoupling high-dimensional fusion features into target muscle group reduction indices and quantitative values ​​of joint mobility limitation. Based on this, it automatically calculates and generates underlying dynamic auxiliary parameters for external intelligent intervention devices, effectively solving the technical problems of homogenization in traditional rehabilitation programs and the difficulty in quantifying the execution of manual interventions.

[0068] This invention constructs a closed-loop feedback control mechanism based on the gradient of temporal characteristic changes. It can adaptively trigger the iterative update of system hardware instructions according to the phased improvement or progressive compensatory needs of the child's function, thereby providing a closed-loop system for the fully automated rehabilitation management of children with DMD, from objective data collection and accurate parameter calculation to dynamic control of external equipment. Attached Figure Description

[0069] Figure 1 This is a flowchart of a rehabilitation management method for children with DMD based on three-dimensional gait analysis combined with NSAA according to the present invention. Detailed Implementation

[0070] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. 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 should fall within the scope of protection of the present invention.

[0071] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but includes other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0072] The present invention will now be described in further detail with reference to the accompanying drawings:

[0073] Example:

[0074] like Figure 1 As shown, a rehabilitation management method for children with DMD based on three-dimensional gait analysis combined with NSAA includes the following steps:

[0075] S100, Obtain basic multimodal feature data: Obtain multidimensional temporal gait feature data of children with DMD and standardized coding vectors based on the Polaris Movement Scale assessment. The multidimensional temporal gait feature data includes spatiotemporal parameters, joint angle data and joint torque data in natural walking state.

[0076] S200, Feature fusion evaluation based on dynamic weighting: Based on the preset feature mapping matrix, the standardized encoding vector and the multidimensional temporal gait feature data are subjected to consistency deviation analysis, and the dynamic weighting mechanism is triggered to perform feature fusion according to the deviation analysis results, and the target muscle group reduction index and joint activity limitation quantification value are output.

[0077] S300, Generate digital intervention and control parameters: Based on the target muscle group decline index and the joint mobility limitation quantification value, calculate and generate digital intervention parameters for external assistive devices, including the dynamic assist parameters of the rehabilitation robot and the structural adjustment parameters of the orthosis;

[0078] S400, Adaptive parameter update based on temporal gradient: Establish a longitudinal database based on the historical assessment sequence of the child, obtain multimodal feature data of the retest after a preset period to calculate the feature change gradient, and adaptively iterate and update the digital intervention parameters based on the feature change gradient.

[0079] Furthermore, the specific implementation process of step S100 is as follows:

[0080] In step S100, the system performs the acquisition and standardized structure transformation of basic multimodal feature data. The system's data input end is equipped with a data interface for interaction with hardware devices and terminals, used to collect objective gait data and subjective scale data of children with DMD in parallel.

[0081] To acquire objective multidimensional temporal gait feature data, the system connects to the three-dimensional spatial motion capture device and the three-dimensional force platform in real time through the underlying communication protocol, and synchronously receives the original continuous gait cycle data containing the spatial coordinate sequence of marker points and the ground reaction force.

[0082] The system's built-in solution module calls the inverse kinematics algorithm to perform rigid body modeling on the spatial coordinate sequence of the marked points. Specifically, based on a preset lower limb biomechanical model, the system establishes local anatomical coordinate systems (LCS) for independent rigid body segments such as the pelvis, thigh, calf, and foot, and obtains the absolute position of each rigid body segment in the global laboratory coordinate system (GCS).

[0083] The system is built based on the homogeneous transformation matrix. A spatial coordinate system transformation model is used to calculate the relative motion between adjacent rigid body segments (such as the thigh and lower leg). For any adjacent proximal rigid body... With the far end rigid body The system extracts the rotation matrices of each component in the global coordinate system. and The relative rotation matrix is ​​calculated using the following matrix multiplication logic. :

[0084]

[0085] In obtaining the relative rotation matrix Then, the system calls a preset Euler Angles decomposition sequence or quaternion solution function to resolve the three-dimensional spatial transformation matrix into scalar angles on three mutually orthogonal planes. Subsequently, the system continuously extracts continuous time steps. Using internal angle scalars, generate physically meaningful three-dimensional joint kinematic sequences. , The grid is composed of curves showing the dynamic range of motion of the child's hip, knee, and ankle joints in the sagittal (flexion and extension), coronal (adduction / abduction), and horizontal (internal and external rotation) planes.

[0086] In the dynamic data analysis layer, the system reads the three-dimensional ground reaction force vector output by the force platform and synchronously inputs it, along with the aforementioned three-dimensional joint kinematic sequence, into the inverse dynamic equation set for iterative solution, thereby calculating the joint dynamic torque sequence. Specifically, the torque sequence includes the hip joint flexion-extension torque sequence, the knee joint flexion-extension torque sequence, and the ankle joint plantar flexion / dorsiflexion torque sequence. The system assigns a unified high-frequency clock timestamp to the above spatiotemporal parameters, kinematic sequences, and dynamic torques, and concatenates them into a structured multidimensional temporal gait feature data matrix.

[0087] For subjective data acquisition, the system executes standardized encoding vector conversion logic based on the Polaris Mobility Scale assessment. The system acquires discrete score data of the child performing 17 preset independent motor function tasks via an interactive terminal. The completion status of each motor function is assigned a preset level label by the clinical input terminal, specifically a discrete integer of 0, 1, or 2. After receiving the one-dimensional raw score array consisting of the 17 level labels, the system triggers the encoding conversion module to perform one-hot encoding processing.

[0088] The specific encoding processing logic is as follows: the system expands and maps each level label with three discrete values ​​into a 3-dimensional binary feature vector (for example, mapping the score "1" to the vector [0,1,0]), thereby mapping the original 17-dimensional integer array into a 51-dimensional sparse matrix. To improve the computational efficiency of subsequent algorithms and eliminate sparse feature redundancy, the system uses Principal Component Analysis (PCA) or an autoencoder network to perform dimensionality reduction mapping on this 51-dimensional sparse matrix, extracting the core principal components whose cumulative variance contribution rate reaches a preset threshold. After the above dimensionality reduction mapping, the system finally outputs a standardized encoding vector with fixed dimensions. This encoding vector has complete machine readability and serves as a digital basis for representing the subjective motor function state of the child.

[0089] After completing the initial construction of the objective time-series data matrix and subjective coding vector, due to the large span of growth and development stages in DMD patients, in order to eliminate the dimensional shift caused by the physical scale differences between individual patients on the gait feature data, the system forcibly triggers data alignment preprocessing logic for physiological scale differences before executing the feature fusion step. The system extracts the basic physiological parameters of the patient's current state from the electronic medical record interface. These physiological parameters strictly include real-time age, height scalar, and weight scalar. The system calls a preset compensation algorithm to construct a scale compensation factor using the above basic physiological parameters. The specific construction and normalization scaling processing logic is as follows: the system sets the weight scalar as the denominator of the dynamic compensation and the height scalar as the denominator of the spatiotemporal feature compensation. Subsequently, the system executes a division operation instruction, using the scale compensation factor containing the height scalar to perform scaling and division processing on the step length vector and gait velocity vector in the spatiotemporal parameter set; similarly, it uses the scale compensation factor containing the weight scalar to perform scaling and division processing on the joint dynamic torque sequence. After eliminating the influence of dimensions through this division operation, the system generates aligned standard multidimensional temporal gait feature data. This preprocessing logic ensures that the data matrix input to the subsequent fusion algorithm is strictly independent of the individual child's physical body size differences in terms of numerical distribution, thereby eliminating pseudo-dynamic decline or improvement caused by changes in height and weight.

[0090] Furthermore, the specific implementation process of step S200 is as follows:

[0091] First, the aligned standard multidimensional temporal gait feature data and standardized encoding vectors are retrieved from memory. Since these two sets of data belong to the objective biomechanical domain and the subjective clinical evaluation domain respectively in terms of underlying data structure and physical dimensions, the system calls a preset feature mapping matrix to execute a cross-domain projection instruction. Specifically, the system constructs the feature mapping matrix based on canonical correlation analysis or a pre-trained autoencoder network, and through linear transformation, forces the standardized encoding vectors and the standard multidimensional temporal gait feature data to be projected into the same high-dimensional hidden semantic feature space, thereby obtaining a fully dimensionally aligned subjective evaluation feature vector. With objective gait feature vector .

[0092] After unifying the feature space, the system triggers a consistency deviation analysis module, designed to quantify the degree of objective conflict between the clinician's visual assessment and the measurement results from the 3D capture device. The system extracts the subjective evaluation feature vector and the objective gait feature vector dimension by dimension. Dimensional components and The processor is then invoked to perform Euclidean space distance calculation, and the global consistency deviation value is calculated using the following error function.

[0093]

[0094] in, The dimension of the feature space. and The objective gait feature vector and the subjective evaluation feature vector are respectively in the th... The components of the dimension. This consistency deviation value. It directly represents the divergence gradient of the subjective and objective data captured by the system under the same feature dimension.

[0095] To mitigate the risk of assessment failure in complex cases using traditional fixed-weighted methods, the system incorporates a dynamic weighting mechanism. The system reads preset basic objective weights. (System default value is 0.6), upper limit of objective weight (The system default value is 0.7), and the bias compensation coefficient is set based on the prior samples. Subsequently, the system will assign the consistency deviation value. Introducing a dynamic weight boundary calculation model, the following operations are performed to generate dynamic objective weights. :

[0096]

[0097] The underlying logic of this operation instruction is: when the subjective and objective data are highly consistent (i.e.) (Approaching zero), the system assigns a baseline weight of 0.6 to objective gait data; once the consistency bias increases significantly (i.e., the subjective scale fails to accurately reflect the child's true compensatory abnormalities), the system will use a product term... Positively increase the weight of objective data in decision-making, and through extreme value functions A hard cutoff threshold of 0.7 is set to prevent over-reliance on a single data source due to sporadic sensor noise. Simultaneously, the system executes subtraction instructions to calculate the corresponding dynamic subjective weights. .

[0098] After completing the adaptive calculation of the two-end weights, the system executes vector-level scalar multiplication and addition instructions to perform weighted fusion of the two sets of feature vectors and calculates the fused feature vector as the output. :

[0099]

[0100] In order to transform the highly abstract fusion feature vector into a digital basis for guiding external rehabilitation equipment, the system uses the fusion feature vector... The data is fed into a pre-defined feature decoupling model. This model employs a deep neural network with a multi-branch output architecture, where the convolutional kernels and fully connected layers are fine-tuned using a DMD proprietary database. The decoupling model separates and extracts feature channels from the fused feature vector, outputting two orthogonal feature subsets in parallel: a muscle weakness feature subset and a joint limitation feature subset.

[0101] For the extracted subset of muscle weakness features, the system performs gradient thresholding on the activation values ​​within it. The system then intersects the activation values ​​of features from different dimensions with a preset muscle function degeneration boundary curve to quantify and output a target muscle group decline index. This index, in the form of a structured array, discretizes the true decline state of the child's hip extensor muscles, knee extensor muscles, or ankle plantar flexor muscles, indicating mild, moderate, or severe decline. Simultaneously, for the subset of joint restriction features, the system retrieves the lower limb skeletal kinematic topology map to perform inverse kinematic space calculations, searching for extreme points of the movement trajectory of the restricted joint site on corresponding anatomical planes such as the sagittal or coronal planes, and extracting the movement limit boundary. Finally, the system generates a high-precision quantified value of joint movement restriction. This quantified value not only includes the three-dimensional coordinates of the labeled restricted anatomical site but also outputs the specific absolute value of the restricted angle (such as the residual value of the maximum dorsiflexion angle of the ankle joint). Thus, the system completes the substantial transformation of multi-source raw data (subjective and objective) into a basis for precise equipment intervention.

[0102] Furthermore, the specific implementation process of step S300 is as follows:

[0103] When the system's local area communication network detects an access signal from the smart ankle-foot orthosis node, the system automatically invokes the static structure adjustment parameter generation thread. The system first parses the joint mobility limitation quantification value input to this thread, and then extracts the child's current maximum ankle dorsiflexion angle during a standard gait cycle through addressing. Subsequently, the system accesses the built-in standard biomechanical database and retrieves the target dorsiflexion angle from healthy children of the same age that match the child's baseline physiological parameters. The system calculates the angle defect compensation amount using subtraction commands. Based on this, the system selects the reduction index specifically corresponding to the ankle plantar flexor muscles from the target muscle group reduction index. Instantiate and substitute into the preset dynamic backbend auxiliary stiffness calculation equation:

[0104]

[0105] During the execution of the above equations, the system reads the factory-set physical reference stiffness of the orthosis. and the set stiffness compensation coefficient The final dynamic backbend auxiliary stiffness is output through multiplication-addition hybrid operations. After completing the scalar calculation, the system will output the stiffness value. The parameters are mapped to the corresponding support phase angle range, serialized into structural adjustment parameters including the number of stepper motor rotation pulses or pneumatic valve limit thresholds, and sent to the microcontroller of the intelligent ankle-foot orthosis via the underlying communication bus.

[0106] When the system detects the connection status of the lower limb rehabilitation exoskeleton robot node, it allocates computing resources to execute a dynamic real-time torque generation thread. Based on the target muscle group reduction index, the system discretizes the level of dynamic functional deficit in the hip and knee extensor muscles of the child. The system's built-in phase oscillator, based on real-time kinematic feedback, divides the child's current gait cycle into discrete auxiliary intervals (e.g., early swing phase, late stance phase) for different phases. Upon entering the corresponding auxiliary interval, the system invokes high-frequency cyclic control statements to continuously calculate the real-time auxiliary torque of the lower limb rehabilitation exoskeleton robot at the target joint at a fixed clock cycle. Its dynamic auxiliary equations are specifically configured as follows:

[0107]

[0108] In this iterative operation logic, the system at each time phase Extracting reference joint torques for standard gait Introducing a nonlinear exponential gain factor Muscle loss index The system incorporates a negative exponential decay term to simulate the static compensatory demand that increases non-linearly with muscle strength decline; simultaneously, the system reads the current angular velocity of the target joint in real time via a sensor feedback interface. and compare it with the preset joint damping coefficient. The static compensation requirement is multiplied to generate a dynamic compensation term. The system then superimposes the static compensation requirement with the dynamic compensation term to output the real-time auxiliary torque. The timing curve is then obtained. Subsequently, the system extracts the extreme points of the curve as the motor trigger thresholds, and compiles the entire timing curve into a low-level control message conforming to the EtherCAT or CANopen protocol format, which is then sent to the servo drive end of the exoskeleton robot as the dynamic auxiliary parameters.

[0109] To address the common respiratory function decline in children with DMD, the system instantiates a parameter generation thread for the intelligent respiratory training device in parallel. The system reverse-engineers the standardized encoding vector generated in step S100, extracting a subset of subjective features specifically characterizing core trunk control through masking. The system applies specific prior weight distributions to each dimension component within this feature subset and performs inner product operations to output a scalarized chest expansion limitation coefficient. Using this limitation coefficient and a preset standard respiratory pressure boundary, the system performs mapping interpolation calculations to directly generate the target positive end-expiratory pressure (PEEP) value and the inspiratory valve opening pressure threshold for the intelligent respiratory training device. The system packages the positive pressure target value, the opening pressure threshold, and the checksum into a standard data frame format and pushes it to the execution end of the intelligent respiratory training device via a wireless communication link, forcibly overwriting its local operating parameter register.

[0110] Furthermore, the specific implementation process of step S400 is as follows:

[0111] The system receives current-phase remeasured multimodal feature data from interactive terminals or external sensors. To ensure consistency in data processing dimensions and units, the system forcibly reuses the underlying processing pipelines of steps S100 to S200, performing isomorphic operations, including spatial coordinate system transformation, scale alignment, and dynamic weighting, on the remeasured multimodal feature data to generate the current fused feature vector. After obtaining the baseline and current sets of high-dimensional feature vectors, the system's built-in comparator module calls vector subtraction and temporal difference instructions. The system extracts corresponding elements between the two sets of vectors channel by channel, calculates the absolute numerical change of the spatiotemporal parameter components, calculates the relative rate of change of the quantized value of joint movement restriction, and uses a matrix concatenation operator to assemble the above calculation results into a gradient vector containing multidimensional improvement or decline trend characteristics.

[0112] After acquiring the changing gradient vector, the system inputs it into the closed-loop feedback control engine for a two-branch threshold determination logic. The first branch corresponds to the positive feedback update logic: the system continuously monitors the joint activity improvement gradient in the changing gradient vector. When the system determines that the improvement gradient reaches a preset positive gain threshold (for example, the system identifies that the gait characteristic component fed back by the sensor jumps above the 0.6 m / s threshold, and the positive increment of the residual value of the maximum dorsiflexion angle of the ankle joint reaches the determination range of 3° to 5°), the system determines that the current hardware assistance intervention is excessive and then triggers the degraded operation logic. The system generates a degraded assistance instruction. At the execution level, the system sends a timing overwrite instruction to the power and timing management chip of the intelligent ankle-foot orthosis, forcibly reducing and rewriting its preset daytime working time cycle from the initial 6 to 8 hours to 4 to 6 hours; or, according to a preset exponential decay step size, lowering the current dynamic assistance parameters of the lower limb rehabilitation exoskeleton robot, thereby forcing the child to recruit their own residual muscle strength.

[0113] The second branch corresponds to the negative compensatory update logic: given the underlying physiological pattern of progressive decline in DMD, the system synchronously monitors the decline gradient of the target muscle group. When the decline gradient is determined to exceed the preset decline risk threshold (i.e., the closed-loop feedback data indicates that the natural progression of the disease has led to insufficient support from the original hardware), the system immediately triggers the compensatory enhancement mechanism. The system calls the reverse instruction generation module to output upgrade auxiliary instructions, directly rewriting the value of the dynamic dorsiflexion auxiliary stiffness register on the control board of the intelligent ankle-foot orthosis through the underlying communication bus, or adding the target damping coefficient and the peak limit of the dynamic auxiliary torque curve to the robot's underlying servo drive node. Through the above-mentioned dual-branch threshold determination and hardware parameter adaptive rewriting mechanism, the system achieves closed-loop iteration and dynamic correction without human intervention, ensuring that the working state of the external auxiliary equipment and the time-varying multi-dimensional functional characteristics of the child maintain strict dynamic coupling.

[0114] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A rehabilitation management method for children with DMD based on three-dimensional gait analysis combined with NSAA, characterized in that, Includes the following steps: S100, Obtain basic multimodal feature data: Obtain multidimensional temporal gait feature data of children with DMD and standardized coding vectors based on the Polaris Movement Scale assessment. The multidimensional temporal gait feature data includes spatiotemporal parameters, joint angle data and joint torque data in natural walking state. S200, Feature fusion evaluation based on dynamic weighting: Based on the preset feature mapping matrix, the standardized encoding vector and the multidimensional temporal gait feature data are subjected to consistency deviation analysis, and the dynamic weighting mechanism is triggered to perform feature fusion according to the deviation analysis results, and the target muscle group reduction index and joint activity limitation quantification value are output. S300, Generate digital intervention and control parameters: Based on the target muscle group decline index and the joint mobility limitation quantification value, calculate and generate digital intervention parameters for external assistive devices, including the dynamic assist parameters of the rehabilitation robot and the structural adjustment parameters of the orthosis; S400, Adaptive parameter update based on temporal gradient: Establish a longitudinal database based on the historical assessment sequence of the child, obtain multimodal feature data of the retest after a preset period to calculate the feature change gradient, and adaptively iterate and update the digital intervention parameters based on the feature change gradient.

2. The rehabilitation management method for children with DMD based on three-dimensional gait analysis combined with NSAA as described in claim 1, characterized in that, The acquisition of multidimensional temporal gait feature data specifically includes: Acquire continuous gait cycle data collected by a 3D spatial motion capture device, and parse the continuous gait cycle data into the following structured temporal features: The spatiotemporal parameter set includes the stride length, stride frequency, and stride speed characteristic sequences of the child when walking naturally in the preset test space; Three-dimensional joint kinematic sequence, including dynamic range curves of the hip, knee and ankle joints of the child in the sagittal, coronal and horizontal planes; And joint dynamic torques, including calculated hip flexion / extension torques, knee flexion / extension torques, and ankle plantar flexion / dorsiflexion torques for the child.

3. The rehabilitation management method for children with DMD based on three-dimensional gait analysis combined with NSAA as described in claim 2, characterized in that, The acquisition of the standardized coding vector based on the Polaris Mobility Scale assessment specifically includes: The discrete score data of the child when performing 17 preset independent motor function tasks were obtained, and the discrete score data was marked with preset level labels according to the degree of functional completion. The discrete scoring data is processed by one-hot encoding or dimensionality reduction mapping to convert the grade labels into machine-readable standardized encoding vectors with fixed dimensions to characterize the child's subjective motor function status.

4. The rehabilitation management method for children with DMD based on three-dimensional gait analysis combined with NSAA as described in claim 3, characterized in that, Prior to the dynamic weighted feature fusion evaluation, a data alignment preprocessing step for physiological scale differences is also included, specifically: Extract the child’s basic physiological parameters, which include at least age, height and weight; Construct a scale compensation factor based on the aforementioned basic physiological parameters; The spatiotemporal parameter set and the joint dynamic torque are normalized and scaled using the scale compensation factor to eliminate the dimensional shift caused by the physical scale differences between different children on the gait feature data, and to obtain aligned standard multidimensional temporal gait feature data.

5. The rehabilitation management method for children with DMD based on three-dimensional gait analysis combined with NSAA according to claim 4, characterized in that, The step of performing consistency deviation analysis on the standardized encoding vector and the multidimensional temporal gait feature data, and triggering a dynamic weighting mechanism for feature fusion based on the deviation analysis results, specifically includes: By projecting the standardized encoding vector onto the same feature space as the standard multidimensional temporal gait feature data using the preset feature mapping matrix, an aligned subjective evaluation feature vector is obtained. With objective gait feature vector ; Calculate the consistency deviation between the subjective evaluation feature vector and the objective gait feature vector. : in, The dimension of the feature space. and The objective gait feature vector and the subjective evaluation feature vector are respectively in the th... Dimensional components; Based on the consistency deviation value Trigger the dynamic weighting mechanism to calculate dynamic objective weights for objective gait data. : in, The preset basic objective weight is set to 0.6; This represents the upper limit of the objective weight, with a value of 0.

7. This is the preset deviation compensation coefficient; Calculate the corresponding dynamic subjective weights This ensures that the value of the dynamic subjective weight is between 0.3 and 0.

4. Based on the calculated dynamic objective weights and dynamic subjective weights, the two sets of feature vectors are weighted and fused to calculate the fused feature vector. : 。 6. The rehabilitation management method for children with DMD based on three-dimensional gait analysis combined with NSAA according to claim 5, characterized in that, The output target muscle group reduction index and joint mobility limitation quantification value specifically include: The fused feature vector is input into a preset feature decoupling model to separate and extract the muscle weakness feature subset and the joint limitation feature subset; Gradient threshold determination is performed on the subset of muscle weakness features, and the target muscle group decline index is quantitatively output. The target muscle group decline index is used to discretize and characterize whether the hip extensor muscle group, knee extensor muscle group or ankle plantar flexor muscle group of the child is in a mild, moderate or severe decline state. The kinematic space inverse solution is performed on the restricted feature subset of the joint to extract the motion limit boundary of the restricted joint on the corresponding anatomical plane, and the joint motion restriction quantization value is generated. The joint motion restriction quantization value includes the specific coordinates of the restricted part and the corresponding restricted angle.

7. A rehabilitation management method for children with DMD based on three-dimensional gait analysis combined with NSAA as described in claim 6, characterized in that, The generation of digital intervention parameters for external assistive devices, when the external assistive device is a smart ankle-foot orthosis, specifically includes: The joint range of motion limitation was analyzed to extract the child's current maximum ankle dorsiflexion angle during the gait cycle. ; Based on a pre-set target dorsiflexion angle for healthy children of the same age With the current maximum ankle dorsiflexion angle Calculate the compensation amount for angular loss; The dynamic dorsiflexion assist stiffness of the intelligent ankle-foot orthosis is calculated using the following formula. : in, The physical reference stiffness of the orthotics. The set stiffness compensation coefficient, The ankle plantar flexor muscle group reduction index is the corresponding index in the target muscle group reduction index. The calculated dynamic backbend auxiliary stiffness The corresponding support phase angle range is converted and generated as the structural adjustment parameters of the intelligent ankle-foot orthosis.

8. A rehabilitation management method for children with DMD based on three-dimensional gait analysis combined with NSAA, as described in claim 7, is characterized in that... The generation of digital intervention parameters for external assistive devices, when the external assistive device is a lower limb rehabilitation exoskeleton robot, specifically includes: Based on the target muscle group decline index, the dynamic functional deficit level of the child's hip extensor muscles and knee extensor muscles is determined. The child's gait cycle is divided into different auxiliary intervals, and the real-time auxiliary torque of the lower limb rehabilitation exoskeleton robot at the target joint is calculated using the following dynamic auxiliary equation. : in, The time phase of the gait cycle, To provide the reference joint torque for the standard gait in the corresponding time phase, The corresponding muscle group reduction index is the index of the joint. It is a non-linear exponential gain factor. This is the preset joint damping coefficient. The real-time joint angular velocity fed back by the sensors of the rehabilitation robot; The real-time auxiliary torque The timing curves and associated motor trigger thresholds are compiled into the low-level control messages of the lower limb rehabilitation exoskeleton robot, which are used as the dynamic auxiliary parameters. The digital intervention parameters also include respiratory impedance parameters for intelligent respiratory training devices, and the specific generation steps include: Extract a subset of subjective features related to core trunk control capabilities from the standardized coding vector; The subjective feature subset is weighted and calculated to evaluate and generate a thoracic expansion restriction coefficient; Based on the chest expansion restriction coefficient, the target value of positive end-expiratory pressure (PEEP) and the opening pressure threshold of the inspiratory valve of the intelligent breathing training device are calculated, and the target positive pressure value and the opening pressure threshold are packaged and sent to the intelligent breathing training device.

9. A rehabilitation management method for children with DMD based on three-dimensional gait analysis combined with NSAA as described in claim 8, characterized in that, The step of obtaining retested multimodal feature data after a preset period to calculate the feature change gradient specifically includes: Extract the baseline fusion feature vector of the child at the initial time phase or the previous assessment cycle from the longitudinal database; The retested multimodal feature data of the current time phase after intervention for at least one preset period is obtained, and the current fused feature vector is generated through the same processing flow as steps S100 to S200. By comparing the current fused feature vector with the benchmark fused feature vector, the absolute changes in spatiotemporal parameters and the relative rate of change of the quantized value of joint activity limitation are extracted respectively, and they are spliced ​​together to form a change gradient vector containing multidimensional improvement trend or decline trend features.

10. A rehabilitation management method for children with DMD based on three-dimensional gait analysis combined with NSAA as described in claim 9, characterized in that, The adaptive iterative update of the digital intervention parameters based on the feature change gradient specifically includes parsing the change gradient vector and executing the following closed-loop update strategy: When it is determined that the joint activity improvement gradient in the changing gradient vector reaches a preset positive gain threshold, a degradation assistance command is generated. The degradation assistance command is used to reduce the current dynamic assistance parameters of the lower limb rehabilitation exoskeleton robot by a preset step size, or to reduce the preset daytime working time cycle of the intelligent ankle-foot orthosis. When the target muscle group's decline gradient in the changing gradient vector exceeds a preset decline risk threshold, a compensatory enhancement mechanism is triggered, automatically generating an upgrade assistance command. The upgrade assistance command is used to increase the dynamic dorsiflexion assist stiffness of the intelligent ankle-foot orthosis or increase the target damping coefficient of the lower limb rehabilitation exoskeleton robot.

Images