Orthopedic rehabilitation intelligent monitoring method

Abstract

The invention provides an intelligent monitoring method for orthopedic rehabilitation, and relates to the technical field of intelligent medical systems. The method comprises the steps of obtaining user attribute data and a user rehabilitation scheme of a to-be-monitored target user; determining predicted stress stimulation data of the orthopedic rehabilitation equipment corresponding to each simulation posture simulation model based on the equipment stimulation evaluation model; matching a target plan time period from the plurality of plan time periods according to the sampling time of the actually measured stress stimulation data, and identifying whether the target user has an abnormal rehabilitation state or not according to the predicted stress stimulation data corresponding to the target plan time period and the actually measured stress stimulation data; and when it is identified that the rehabilitation abnormal state exists, sending an early warning notification to a preset medical care terminal according to the actually measured stress stimulation data of the target user. Therefore, in combination with predicted and actually measured stress stimulation data, scientific assessment and personalized rehabilitation guidance on rehabilitation actions are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of smart medical systems, and in particular to an intelligent monitoring method and system for orthopedic rehabilitation. Background Art

[0002] With the continuous advancement of medical technology, the field of orthopedic rehabilitation has achieved remarkable development. At present, there are some rehabilitation assistive devices on the market, such as walking crutches and foot pressure detection pads, but these devices generally only provide basic physical support and some simple detection functions, and cannot fully record the patient's kinematic characteristics.

[0003] For orthopedic patients, especially those with lower limb fractures, scientific and reasonable mechanical stimulation is critical in postoperative rehabilitation training. Mechanical stimulation has a positive effect on bone healing, and appropriate mechanical stimulation can promote the growth and repair of bone cells. However, complete restriction of movement is also disadvantageous, because the lack of appropriate mechanical stimulation can lead to decreased bone density and delayed recovery.

[0004] Existing rehabilitation equipment is difficult to record and analyze patients' kinematic characteristics and dynamic changes of mechanical stimulation in real time. However, during rehabilitation training, patients may suffer from nonunion or delayed healing of fractures due to excessive mechanical stimulation, or their rehabilitation progress may be affected due to insufficient mechanical stimulation. Therefore, the lack of real-time monitoring and feedback mechanism leads to certain limitations.

[0005] In addition, most existing rehabilitation programs are designed for general use and fail to fully consider the individual differences of different patients. The rehabilitation process and needs of each patient are different, and personalized rehabilitation guidance is particularly important for improving rehabilitation results. Existing equipment lacks the ability to dynamically adjust rehabilitation programs based on real-time monitoring data and cannot provide personalized rehabilitation guidance.

[0006] In response to the above problems, the industry has not yet proposed a better technical solution. Summary of the invention

[0007] The embodiments of the present invention provide an intelligent monitoring method and system for orthopedic rehabilitation, which are used to at least solve the problem that current rehabilitation equipment cannot implement personalized rehabilitation guidance because it cannot record and analyze the patient's kinematic characteristics and dynamic changes of mechanical stimulation in real time.

[0008] In a first aspect, an embodiment of the present invention provides an intelligent monitoring method for orthopedic rehabilitation, the method comprising: obtaining user attribute data and a user rehabilitation plan of a target user to be monitored; the user rehabilitation plan comprises a plurality of consecutive planned time periods and corresponding simulated posture simulation models; each of the planned time periods has a corresponding rehabilitation action, and the simulated posture simulation model is a simulation model that simulates the target user when performing the corresponding rehabilitation action; for each simulated posture simulation model, the user attribute data, the simulated posture simulation model and the rehabilitation action are input into a preset device stimulation evaluation model to determine predicted stress stimulation data of an orthopedic rehabilitation device corresponding to the simulated posture simulation model; the predicted stress stimulation data represents predicted stress stimulation data that will be generated by the target user on the orthopedic rehabilitation device by performing rehabilitation actions; when it is detected that the orthopedic rehabilitation device samples measured stress stimulation data, matching a target planned time period from the plurality of planned time periods according to the sampling time of the measured stress stimulation data, and identifying whether the target user has an abnormal rehabilitation state according to the predicted stress stimulation data corresponding to the target planned time period and the measured stress stimulation data; when it is identified that an abnormal rehabilitation state exists, sending an early warning notification to a preset medical terminal according to the measured stress stimulation data of the target user.

[0009] In a second aspect, an embodiment of the present invention provides an orthopedic rehabilitation intelligent monitoring system, the system comprising: an acquisition unit, for acquiring user attribute data and a user rehabilitation plan of a target user to be monitored; the user rehabilitation plan comprises a plurality of consecutive planned time periods and corresponding simulated posture simulation models; each of the planned time periods has a corresponding rehabilitation action, and the simulated posture simulation model is a simulation model that simulates the target user when performing the corresponding rehabilitation action; a stimulation evaluation unit, for inputting the user attribute data, the simulated posture simulation model and the rehabilitation action into a preset device stimulation evaluation model for each simulated posture simulation model, so as to determine the prediction of the orthopedic rehabilitation device corresponding to the simulated posture simulation model. stress stimulation data; the predicted stress stimulation data represents the predicted stress stimulation data that will be generated by the target user on the orthopedic rehabilitation device through the implementation of rehabilitation actions; a rehabilitation abnormality identification unit, which is used to match the target planned time period from the multiple planned time periods according to the sampling time of the measured stress stimulation data when it is detected that the orthopedic rehabilitation device samples the measured stress stimulation data, and to identify whether the target user has an abnormal rehabilitation state according to the predicted stress stimulation data corresponding to the target planned time period and the measured stress stimulation data; an early warning unit, which is used to send an early warning notification to a preset medical terminal according to the measured stress stimulation data of the target user when an abnormal rehabilitation state is identified.

[0010] In a third aspect, an embodiment of the present invention provides an electronic device, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the steps of the above method.

[0011] In a fourth aspect, an embodiment of the present invention provides a storage medium, in which one or more programs including execution instructions are stored. The execution instructions can be read and executed by an electronic device (including but not limited to a computer, a server, or a network device, etc.) to execute the steps of the above-mentioned method of the present invention.

[0012] In a fifth aspect, an embodiment of the present invention further provides a computer program product, wherein the computer program product comprises a computer program stored on a storage medium, wherein the computer program comprises program instructions, and when the program instructions are executed by a computer, the computer executes the steps of the above method.

[0013] The orthopedic rehabilitation intelligent monitoring method, system, electronic device and non-transitory computer-readable storage medium provided by the present invention can produce at least the following technical effects:

[0014] (1) By obtaining the user attribute data and user rehabilitation plan of the target user, the patient's kinematic characteristics can be comprehensively recorded and analyzed, and the multidimensional data can be input into the preset device stimulation evaluation model. The deep learning model can accurately predict the stress stimulation data that should be generated in different stages of the rehabilitation process for users of different body types, providing a basis for realizing personalized rehabilitation data recording and analysis.

[0015] (2) By applying deep learning technology, the predicted stress stimulation data of each simulated posture simulation model is calculated, thereby realizing the prediction of stress stimulation in different rehabilitation time periods. In addition, by combining the predicted stress stimulation data with the measured stress stimulation data, the changes in mechanical stimulation during the rehabilitation process can be dynamically evaluated to ensure that the user's rehabilitation progress is in line with the rehabilitation plan, which helps promote the growth and repair of bone cells.

[0016] (3) By detecting the measured stress stimulation data sampled by the orthopedic rehabilitation equipment and comparing it with the predicted stress stimulation data, it is possible to monitor the abnormal rehabilitation status of the target user at each continuous rehabilitation plan time in real time. When an abnormal rehabilitation status is detected, the system will immediately send an early warning notification to the preset medical terminal, realizing a real-time monitoring and feedback mechanism and personalized rehabilitation guidance. For example, it can effectively avoid rehabilitation problems caused by excessive or insufficient mechanical stimulation, such as fracture nonunion or delayed rehabilitation, and improve the safety and effectiveness of rehabilitation.

[0017] Through this technical solution, the individual differences of users are comprehensively considered, and the rehabilitation plan is dynamically adjusted by using user attribute data and simulated rehabilitation postures. According to the different rehabilitation processes and body characteristics of different patients, a simulated posture simulation model is used, combined with predicted and measured stress stimulation data, to achieve scientific evaluation of rehabilitation movements and personalized rehabilitation guidance, thereby improving rehabilitation effects and patient satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0019] Figure 1 A flowchart showing an example of an intelligent monitoring method for orthopedic rehabilitation according to an embodiment of the present invention is shown;

[0020] Figure 2 An operation flow chart showing an example of identifying abnormal rehabilitation status of a target user according to an embodiment of the present invention;

[0021] Figure 3 A schematic diagram showing a structural connection of an example of a device stimulation evaluation model according to an embodiment of the present invention is shown;

[0022] Figure 4 It shows an operation flow chart of an example of obtaining a user rehabilitation plan according to an embodiment of the present invention;

[0023] Figure 5 A schematic diagram showing a structural connection of an example of a posture conversion network according to an embodiment of the present invention is shown;

[0024] Figure 6 A structural block diagram of an example of an intelligent monitoring system for orthopedic rehabilitation according to an embodiment of the present invention is shown;

[0025] Figure 7 It is a schematic structural diagram of an embodiment of an electronic device of the present invention. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical solution and advantages of the embodiment of the present invention clearer, the technical solution of the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings of the embodiment of the present invention. Obviously, the described embodiment is a part of the embodiment of the present invention, not all of the embodiments. Based on the described embodiment of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0027] Unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. Similarly, words such as "one", "one" or "the" do not indicate a quantitative limitation, but indicate the existence of at least one. Words such as "include" or "comprise" mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. Words such as "connect" or "connected" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect.

[0028] It should be noted that the terms "up", "down", "left", "right", "front", "back", etc. used in the present invention are only used to indicate relative position relationships. When the absolute position of the object being described changes, the relative position relationship may also change accordingly.

[0029] It should be noted that during rehabilitation training, early comprehensive movement restriction can prevent displacement of the fracture site, but as the rehabilitation process progresses, appropriate increase in exercise and mechanical stimulation will help bone healing and functional recovery. However, excessive mechanical stimulation may cause nonunion or delayed healing of fractures. Therefore, during rehabilitation training, real-time understanding and monitoring of the patient's mechanical stimulation status is crucial to the rehabilitation effect.

[0030] Figure 1 A flowchart of an example of an intelligent monitoring method for orthopedic rehabilitation according to an embodiment of the present invention is shown.

[0031] Regarding the execution subject of the method of the embodiment of the present invention, it can be any controller or processor with computing or processing capabilities, which can be integrated or partially integrated into a server or a coronary heart disease risk prediction platform, so as to significantly improve the scientificity, effectiveness and safety of orthopedic rehabilitation training by introducing comprehensive recording and analysis, real-time monitoring and feedback, personalized rehabilitation guidance and other technical means. The deep learning model is used to predict and identify the dynamic changes of mechanical stimulation during the rehabilitation process, providing patients with a more accurate and personalized rehabilitation plan, promoting the effective healing of fractures and improving the overall rehabilitation effect.

[0032] In some examples, it can be integrated into the configuration server by software, hardware, or a combination of software and hardware, which should not be limited here. It should be understood that one or more steps involved in the following process can be implemented by one or more controllers or software installed and deployed in the client or server.

[0033] like Figure 1As shown, in step S110, user attribute data and a user rehabilitation plan of a target user to be monitored are obtained, and the user rehabilitation plan includes a plurality of consecutive planned time periods and corresponding simulated posture simulation models.

[0034] Here, the user attribute data includes user information, user weight, body mass index, waist-to-hip ratio, body circumference measurement and user gender. Each planned time period has corresponding rehabilitation actions. The simulated posture simulation model is a simulation model that simulates the target user when performing the corresponding rehabilitation actions.

[0035] In some embodiments, the basic information of the target user is obtained through a medical system or user self-report, including user information (such as age, gender, medical history, etc.), weight, body mass index (BMI), waist-to-hip ratio, body circumference measurement (such as waist, chest and hip circumference) and other data. In addition, a professional medical team can formulate a user rehabilitation plan based on the user's rehabilitation needs and medical advice, which is divided into multiple planned time periods (such as daily or weekly rehabilitation training plans), and each time period contains specific rehabilitation actions. It should be noted that the rehabilitation actions in different time periods can be the same set of actions or different actions, and all belong to the example scope of the present invention. The specific principle is that, on the one hand, the device stress stimulation distribution when the user performs different actions is detected to detect the user's injury recovery progress (for example, the recovery of lower limb injuries). On the other hand, the device stress stimulation distribution can also be adjusted due to changes in the user's recovery stage when the user performs the same action. For example, the stress stimulation distribution of the smart foot pad will also change accordingly due to the stepping action in different recovery stages of ankle injuries. Furthermore, the rehabilitation movements in the rehabilitation program can be used in combination with user attribute data to generate a corresponding simulated posture simulation model, which can be achieved through various computer simulation technologies to simulate personalized posture models of users with different body shapes when performing rehabilitation movements.

[0036] In step S120, for each simulated posture simulation model, user attribute data, simulated posture simulation model and rehabilitation action are input into a preset device stimulation evaluation model to determine the predicted stress stimulation data of the orthopedic rehabilitation device corresponding to the simulated posture simulation model.

[0037] In some embodiments, the device stimulation evaluation model adopts various deep learning models, such as LSTM (Long Short-Term Memory), CNN (Convolutional Neural Network) or Transformer model, etc. Specifically, after the device stimulation evaluation model is trained and converged through a large amount of sample data, the user attribute data, the simulated posture simulation model and the rehabilitation action are input into the device stimulation evaluation model to process and analyze the input data and calculate the predicted stress stimulation data corresponding to each simulated posture.

[0038] Here, the predicted stress stimulation data indicates the predicted stress stimulation data that the target user will generate on the orthopedic rehabilitation device by performing rehabilitation actions. Specifically, if the user recovers normally, when performing corresponding rehabilitation actions at a specific stage, the stress stimulation data generated on the orthopedic rehabilitation device should be not much different from the predicted stress stimulation data.

[0039] It should be noted that the rehabilitation action can be a specific action or a coherent set of actions, and the corresponding predicted stress stimulation data can be a single-valued or multi-valued sequence. In this case, the predicted stress stimulation data can also be single-valued or time series data.

[0040] In step S130, when it is detected that the orthopedic rehabilitation device samples the measured stress stimulation data, the target planned time period is matched from multiple planned time periods according to the sampling time of the measured stress stimulation data, and it is identified whether the target user has an abnormal rehabilitation state based on the predicted stress stimulation data corresponding to the target planned time period and the measured stress stimulation data.

[0041] In some examples of embodiments of the present invention, the orthopedic rehabilitation equipment includes any one of the following: smart insoles, smart crutches and smart protective gear, and the stimulation type of the stress stimulation data includes at least one of the following: compression stress, shear stress, tensile stress, torsional stress and impact stress. By real-time monitoring of the pressure distribution and stress changes of the foot through smart insoles, the stress distribution during gait and walking can be evaluated to help adjust the rehabilitation training program. Force sensors and acceleration sensors can be equipped in smart crutches to monitor the stress and motion state during use, and can assist walking training and provide real-time feedback. The stress and activity of the injured part are monitored through smart protective gear, and sensors are equipped to detect stress changes during rehabilitation training, thereby protecting the injured part and providing real-time monitoring data.

[0042] Here, compressive stress defines the pressure in the vertical direction, which is usually generated when standing and walking, and is used to evaluate the stress of the foot and lower limbs under weight-bearing conditions. Shear stress defines the stress parallel to the contact surface, which is usually generated during movement and sliding. It is used to analyze the shear force distribution of the foot and lower limbs during gait and movement. Tensile stress defines the pulling force along the axis direction, and the stress is concentrated on the muscles and ligaments. It is used to evaluate the force of muscles and ligaments in rehabilitation training. Torsional stress defines the stress caused by rotational movement acting on joints and bones, which is used to analyze the stress distribution and changes in joint mobility training. Impact stress defines the stress applied quickly in a short period of time, such as the impact force during jumping and running, which is used to evaluate the stress in dynamic movement. Therefore, by combining a variety of rehabilitation equipment and stress stimulation types, it is possible to comprehensively monitor the stress of users in rehabilitation training and achieve effective and accurate stress anomaly detection.

[0043] In some embodiments, when a user wears an orthopedic rehabilitation device to perform rehabilitation movements, the sensor in the orthopedic rehabilitation device can measure and generate measured stress stimulation data in real time and upload it to the orthopedic rehabilitation intelligent monitoring platform. At this time, the system extracts the sampling time of the measured stress stimulation data and compares it with the planned time periods of multiple rehabilitation stages to obtain a matching target planned time period. Then, the predicted stress stimulation data of the target planned time period is compared with the measured stress stimulation data to identify whether the target user has an abnormal rehabilitation state.

[0044] Figure 2 FIG. 1 is a flowchart showing an example of identifying the abnormal state of rehabilitation of a target user according to an embodiment of the present invention. Figure 2 As shown, in step S210, the stress deviation value between the predicted stress stimulation data and the measured stress stimulation data is calculated. In step S220, when it is detected that the stress deviation value exceeds the preset stress deviation threshold, it is determined that the target user has an abnormal state of rehabilitation. In some embodiments, according to clinical research and expert opinions, a reasonable stress deviation threshold is set, which can be a static value or a dynamically adjusted value to adapt to different rehabilitation stages and individual differences, and the abnormal state can also be divided into multiple types (such as mild abnormality, moderate abnormality and severe abnormality). Thus, by accurately calculating the deviation value between the predicted stress and the measured stress, the system can accurately monitor the actual effect of each action in the rehabilitation process and can identify the abnormal state of rehabilitation at the first time.

[0045] In step S140, when an abnormal rehabilitation state is identified, an early warning notification is sent to a preset medical terminal according to the measured stress stimulation data of the target user.

[0046] In some embodiments, when the system identifies an abnormal state of rehabilitation, the early warning mechanism is immediately triggered, and a detailed early warning notification is generated based on the measured stress stimulation data of the target user, and sent to the medical terminal through a preset communication channel (such as mobile phone text messages, emails, or medical system notifications). In this way, medical staff can intervene and adjust the user's original planned rehabilitation plan in a timely manner based on the content of the early warning notification, which can contain detailed information about the abnormal state, such as the type of abnormality, time, and the difference between the measured data and the predicted data, so as to achieve high precision of personalized medical rehabilitation plans.

[0047] In terms of business scenarios, the intelligent orthopedic rehabilitation monitoring system of this embodiment can be deployed on the mechanical detection insoles and smart crutches of both feet to comprehensively record the patient's movement status and mechanical stimulation data, and communicate with doctors in a timely manner through remote communication to optimize rehabilitation training plans, thereby accelerating fracture healing and reducing unnecessary complications and rehabilitation delays.

[0048] Figure 3 A schematic diagram of structural connections of an example of a device stimulation evaluation model according to an embodiment of the present invention is shown.

[0049] like Figure 3 As shown, the device stimulation evaluation model 300 adopts a multimodal fusion prediction model based on ensemble learning and self-attention mechanism, which includes an input layer 310, a base learning module 320, a feature fusion layer 330 and an output layer 340. The base learning module 320 includes a random forest layer 321, a bidirectional long short-term memory network layer 322 and a convolutional neural network layer 323.

[0050] More specifically, the input 310 layer is used to receive user attribute data, simulated posture simulation models, and rehabilitation actions.

[0051] In some embodiments, for rehabilitation action sequences, anomaly detection algorithms (such as isolation forests and local anomaly factors) can be used to remove noise and outliers in the data to ensure data accuracy. In addition, user attribute data, time series data, and image data can be standardized to scale data of different scales to a uniform range for ease of model processing and fusion. In addition, the pressure distribution image can be rotated, translated, and scaled to increase data diversity.

[0052] The random forest layer 321 is used to perform integrated prediction processing on user attribute data through multiple decision trees to extract static user attribute feature representation Y RF .

[0053] Here, user attribute data is processed by random forest to extract static features. By integrating multiple decision trees, the model extracts static features and can combine them with other dynamic features for comprehensive analysis, thereby improving the overall prediction performance.

[0054] The bidirectional long short-term memory network layer 322 is used to capture the time dependency corresponding to the rehabilitation action to obtain the corresponding time series feature representation Y Bi-LSTM :

[0055] Forward LSTM processing:

[0056]

[0057] Backward LSTM processing:

[0058]

[0059] Merge the hidden state of the forward LSTM and the hidden state of the backward LSTM to form the output feature representation of the bidirectional LSTM:

[0060]

[0061] In the formula, the rehabilitation action is obtained through the time series data S = [s1, s2, ..., s T ] is defined as follows, T represents the total number of time steps included in the rehabilitation action; s t is the rehabilitation action feature vector at the tth time step; represents the hidden state of the forward LSTM at the tth time step, represents the hidden state of the forward LSTM at the t-1th time step; represents the hidden state of the backward LSTM at the tth time step, represents the hidden state of the backward LSTM at the t+1th time step.

[0062] In this way, by processing the time series data of rehabilitation movements through Bi-LSTM and capturing the time dependency, the forward and backward dependencies of the time series can be considered at the same time, thus improving the modeling ability of the sequence data. By capturing the forward and backward dependencies of the time series data of rehabilitation movements through Bi-LSTM, the model can more accurately understand the dynamic changes in the time series, thereby improving the prediction accuracy of the changes in mechanical stimulation during the rehabilitation process.

[0063] The convolutional neural network layer 323 is used to extract the model space features corresponding to the simulated posture simulation model through the residual network structure:

[0064] The residual network structure contains multiple residual blocks, each of which is implemented through a jump connection:

[0065] Y res =F(X res )+X res , Formula (4)

[0066] Where Yres represents the output of the residual block, X res Represents the input of the residual block, and F represents the combined operation of convolution, batch normalization, and activation function.

[0067] Here, the residual network is used to effectively solve the gradient vanishing problem in the deep network, and improve the ability and accuracy of feature extraction, so that the spatial characteristics of rehabilitation posture can be accurately captured, thereby improving the prediction accuracy of stress stimulation data.

[0068] The feature fusion layer 330 is used to perform weighted fusion on the output features of the base learning modules to obtain the corresponding deep fusion features:

[0069] F=α1Y CNN +α2Y Bi-LSTM +α3Y RF , Formula (5)

[0070] In the formula, F represents the deep fusion feature, Y CNN Represents the model space features of the final output of the convolutional neural network layer; α1, α2, α3 represent adaptive fusion weights, which can be dynamically adjusted through the back-propagation algorithm to ensure that the importance of different features is reasonably reflected in the fusion process.

[0071] Here, the adaptive fusion weights α1, α2, and α3 can be dynamically adjusted through the back-propagation algorithm to ensure that the importance of different features is reasonably reflected in the fusion process. By integrating spatial features, time series features, and static features through the adaptive fusion layer, the model can fully utilize the advantages of different modal data to improve the comprehensiveness and accuracy of the prediction. The fusion weights are dynamically adjusted through back-propagation during the training process to ensure the prominent representation of important features, so that the model can better cope with complex rehabilitation scenarios.

[0072] The output layer 340 is used to perform nonlinear transformation on the deep fusion features through the fully connected layer to predict the stress stimulation data of the output simulated posture simulation model:

[0073] Z=σ(W1F+b1), Formula (6)

[0074] Y pred =σ(W2Z+b2), Formula (7)

[0075] In the formula, σ represents the ReLU activation function, Y pred Represents the predicted stress stimulation data of the simulated posture simulation model; W1, b1 represent the weight matrix and bias term of the first fully connected layer, respectively; W2, b2 represent the weight matrix and bias term of the second fully connected layer, respectively; Z represents the feature representation of the first fully connected layer after nonlinear transformation through the first fully connected layer.

[0076] Here, by integrating user attribute data, simulated rehabilitation posture and rehabilitation movement data, the model can accurately predict stress stimulation based on the personalized data of specific patients. The personalized prediction data provides a scientific basis for rehabilitation doctors, which helps to formulate more accurate rehabilitation training plans and improve rehabilitation effects.

[0077] Through this embodiment, the model design based on multimodal data fusion and adaptive fusion layer can adapt to different types of rehabilitation data and scenarios. The device stimulation evaluation model can achieve high precision, multimodal fusion, personalized guidance, real-time monitoring and feedback, improved robustness and stability, and good scalability and versatility in the scenario of orthopedic rehabilitation intelligent monitoring technology.

[0078] In some examples of embodiments of the present invention, the device stimulation evaluation model uses a comprehensive loss function that combines Mean Squared Error (MSE) loss and Dynamic Time Warping (DTW) loss. MSE loss is used to measure the difference between the predicted stress stimulation data and the real stress stimulation data, and DTW loss is used to calculate the matching error between the predicted stress stimulation data and the real stress stimulation data in time.

[0079] More specifically, the comprehensive loss function L 综合 It is expressed by the following formula:

[0080] L 综合 =λ1L DTW +λ2L MSE , Formula (8)

[0081]

[0082] Where, L DTW and L MSE They represent the dynamic time warping loss and mean square error loss respectively. λ1 and λ2 represent the weight parameters of the corresponding loss terms, which can be set according to business needs or updated through back propagation during training. N represents the total number of samples in the data sample set, Y pred,i represents the predicted stress stimulation data of the i-th sample in the data sample set, Y true,i represents the true stress stimulation data of the i-th sample in the data sample set, Y true,j represents the real stress stimulation data of the jth sample in the data sample set; δ(i,j) is the path constraint condition to ensure the rationality of the pairing of the time series, which indicates whether the path corresponding to data sample i and data sample j is allowed at the corresponding time step; d(Y pred,i ,Y true,j ) indicates Y pred,iand Y true,j The Euclidean distance between .

[0083] It should be noted that the path constraint δ(i,j) ensures that the selected path meets the actual constraints of time series alignment and guarantees the order consistency of time series alignment, that is, the previous time step i corresponds to the previous time step j, and the following time step i+1 corresponds to the following time step j+1, ensuring that the path is continuous in the time series and avoiding jumps or reverse pairings. More specifically, if the pairing (i,j) is allowed (i.e., it meets the path constraint), then δ(i,j) = 1, otherwise δ(i,j) = 0.

[0084] Through this embodiment, based on the MSE loss term, the difference between the predicted stress stimulation data and the actual measured data can be accurately quantified. By minimizing the loss function, the model can accurately predict the stress stimulation data of the target user during the rehabilitation process. Based on the DTW loss term, the time difference between the predicted and actual data can be effectively aligned, and the real dynamic changes of stress stimulation can be captured even when there are inconsistencies in the time scale, which is particularly important for dealing with time lags and variations that may occur in actual rehabilitation training. In addition, the weight parameters are combined with the MSE loss term and the DTW loss term to provide a comprehensive and efficient evaluation mechanism for the orthopedic rehabilitation intelligent monitoring system, which can not only accurately predict and monitor the user's stress stimulation data, but also detect and optimize abnormal situations in the rehabilitation process in real time, provide users with scientific and personalized rehabilitation guidance, and improve the intelligence and practicality of the overall system.

[0085] Figure 4 An operational flowchart of an example of obtaining a user rehabilitation plan according to an embodiment of the present invention is shown.

[0086] like Figure 4 As shown, in step S410, input information is obtained, and the user attribute data and user rehabilitation plan of the target user to be monitored corresponding to the input information are parsed. The user rehabilitation plan includes multiple consecutive planned time periods and corresponding planned rehabilitation action names.

[0087] In some embodiments, the doctor can input various types of information through the system interface. Specifically, various types of user attribute information are obtained through the input information, including the user's basic information, weight, body mass index, waist-to-hip ratio, body circumference measurement, gender, etc. In addition, by parsing the acquired input information, the user rehabilitation plan is extracted, which includes a plurality of consecutive planned time periods or rehabilitation stages, and each planned time period has a corresponding planned rehabilitation action name.

[0088] In step S420, target standard posture simulation models that match the names of the planned rehabilitation movements are determined from the posture model library.

[0089] Here, the posture model library contains multiple rehabilitation action names and corresponding standard posture simulation models. More specifically, each standard posture simulation model has been verified by a professional medical team to ensure its scientificity and accuracy. Therefore, through the posture model library, it is possible to quickly match the corresponding standard posture simulation models for each stage based on the names of each planned rehabilitation action in the user's rehabilitation plan.

[0090] In step S430, for each target standard posture simulation model, the target standard posture simulation model and user attribute data are input into a posture conversion network to convert and generate a corresponding simulated posture simulation model.

[0091] In some embodiments, a pre-trained posture conversion network is used to input the target standard posture simulation model and user attribute data into the network. The posture conversion network can be a model based on deep learning, such as a generative adversarial network or a variational autoencoder, etc. In this way, the posture conversion network converts the standard posture simulation model according to the user's individual attributes (such as weight, height, body circumference, etc.) to generate a simulated posture simulation model suitable for the target user as the simulated posture three-dimensional data of personalized rehabilitation movements.

[0092] In step S440, a user rehabilitation plan is generated according to a plurality of consecutive planned time periods and corresponding simulated posture simulation models.

[0093] In this way, the standard posture simulation model is adjusted in combination with the user's attribute information and rehabilitation demand goals to obtain the simulated posture simulation model of the corresponding rehabilitation movements in different stages, ensuring that the simulated posture simulation model of each planned time period can match the user's personalization in the corresponding rehabilitation stage and ensure the scientificity and effectiveness of the rehabilitation process.

[0094] Figure 5 A schematic diagram of structural connections of an example of a posture conversion network according to an embodiment of the present invention is shown.

[0095] like Figure 5 As shown, the posture conversion network 500 adopts a conditional generative adversarial network, which includes an input layer 510, a posture encoder 520, an attribute encoder 530, a feature fusion layer 540, a generator 550 and a discriminator 560.

[0096] Here, the generator 550 is used to generate a personalized simulated posture simulation model, and the discriminator 560 is used to evaluate the quality of the simulated posture evaluation simulation model during the training phase to ensure the authenticity and reliability of the generated simulated posture simulation model. The discriminator will not be used in the actual reasoning phase. Using the generator and discriminator adversarial training mechanism of cGAN (Conditional Generative Adversarial Network, conditional generative adversarial network), the generator can generate high-quality simulated posture simulation models, while the discriminator continuously optimizes the authenticity of the generated results. This adversarial mechanism significantly improves the accuracy of posture prediction and ensures that the generated posture simulation model has a high degree of authenticity and accuracy.

[0097] More specifically, the input layer 510 is used to receive the target standard posture simulation model P std And user attribute data U.

[0098] The posture encoder 520 is used to extract the standard posture coding features corresponding to the target standard posture:

[0099] F p1 =σ(Conv(P std )), Formula (11)

[0100] F p2 =Pool(F p1 ), Formula (12)

[0101] F p =σ(W p ·F p2 +b p ), Formula (13)

[0102] In the formula, Conv represents the convolution operation, σ represents the ReLU activation function, and F p1 represents the posture convolution feature; Pool represents the pooling operation, F p2 Represents the features of the pooling layer output; W p and b p Denote the weight matrix and bias term of the fully connected layer of the pose encoding, F p Represents standard posture encoding features.

[0103] The attribute encoder 530 is used to extract the user attribute encoding features corresponding to the user attribute data:

[0104] F u =σ(W u ·U+b u ), Formula (14)

[0105] Where W u and bu Represent the weight matrix and bias term of the attribute encoding fully connected layer, F u Represents the user attribute encoding feature.

[0106] By encoding and extracting features from the input data, the posture conversion network can effectively process and filter noisy data and improve the robustness of the model. In addition, the discriminator in the cGAN structure can detect anomalies in the generated model, thereby ensuring the stability and reliability of the generated posture simulation model under different environments and conditions.

[0107] The feature fusion layer 540 is used to adaptively fuse the standard posture coding features and the user attribute coding features to determine the corresponding multimodal fusion features:

[0108] F fused =γ·F p +(1-γ)·F u , Formula (15)

[0109] In the formula, γ represents the adaptive fusion weight, F fused Represents multimodal fusion features.

[0110] Through this embodiment, the standard posture simulation model and user attribute data are adaptively fused, and the posture conversion network can generate a personalized simulated posture simulation model, which can better adapt to the specific rehabilitation needs of each user and improve the scientificity and effectiveness of the rehabilitation program. In addition, the adaptive weight adjustment mechanism in the feature fusion layer can dynamically adjust the weights of posture features and user features according to training data. By continuously optimizing the weight parameters, the model can better capture and reflect the individual differences of users, thereby improving the accuracy and adaptability of the generated model.

[0111] The generator 550 is used to generate a personalized simulated posture simulation model according to the multimodal fusion features:

[0112] F g1 =σ(W g1 ·F fused +b g1 ), Formula (16)

[0113] P sim =Deconv(F g1 ), Formula (17)

[0114] Where W g1 and b g1 Represent the weight matrix and bias term of the fully connected layer of the generator respectively; F g1 represents the high-dimensional representation of the generator; Deconv represents the deconvolution operation, P sim Represents the simulated posture simulation model.

[0115] Through this embodiment, the posture conversion network can simultaneously process and fuse multiple data sources, including standard posture simulation models and user attribute data. Based on multimodal data fusion, it can make full use of information from different data sources to provide more comprehensive and accurate posture prediction and rehabilitation program evaluation. Therefore, based on the generated high-precision and personalized simulated posture simulation model, it helps to improve the accuracy of subsequent stimulation data prediction results, and can provide users with more personalized, scientific and accurate rehabilitation guidance.

[0116] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity of description, they are all expressed as a series of combined actions, but those skilled in the art should be aware that the present invention is not limited by the described order of actions, because according to the present invention, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention. In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0117] Figure 6 A structural block diagram of an example of an orthopedic rehabilitation intelligent monitoring system according to an embodiment of the present invention is shown.

[0118] like Figure 6 As shown, the orthopedic rehabilitation intelligent monitoring system 600 includes an acquisition unit 610 , a stimulation evaluation unit 620 , a rehabilitation abnormality identification unit 630 and an early warning unit 640 .

[0119] The acquisition unit 610 is used to obtain user attribute data and user rehabilitation plan of the target user to be monitored; the user rehabilitation plan includes multiple consecutive planned time periods and corresponding simulated posture simulation models; each of the planned time periods has corresponding rehabilitation actions, and the simulated posture simulation model is a simulation model that simulates the target user when performing the corresponding rehabilitation actions.

[0120] The stimulation evaluation unit 620 is used to input the user attribute data, the simulated posture simulation model and the rehabilitation action into a preset device stimulation evaluation model for each simulated posture simulation model, so as to determine the predicted stress stimulation data of the orthopedic rehabilitation device corresponding to the simulated posture simulation model; the predicted stress stimulation data represents the predicted stress stimulation data that the target user will generate on the orthopedic rehabilitation device by performing rehabilitation actions.

[0121] The rehabilitation abnormality identification unit 630 is used to match the target planned time period from the multiple planned time periods according to the sampling time of the measured stress stimulation data when it is detected that the orthopedic rehabilitation device samples the measured stress stimulation data, and to identify whether the target user has an abnormal rehabilitation state based on the predicted stress stimulation data corresponding to the target planned time period and the measured stress stimulation data.

[0122] The early warning unit 640 is used to send an early warning notification to a preset medical terminal according to the measured stress stimulation data of the target user when an abnormal rehabilitation state is identified.

[0123] In some embodiments, an embodiment of the present invention provides a non-volatile computer-readable storage medium, which stores one or more programs including execution instructions, and the execution instructions can be read and executed by an electronic device (including but not limited to a computer, a server, or a network device, etc.) to execute the above-mentioned orthopedic rehabilitation intelligent monitoring method of the present invention.

[0124] In some embodiments, an embodiment of the present invention further provides a computer program product, which includes a computer program stored on a non-volatile computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer executes the above-mentioned orthopedic rehabilitation intelligent monitoring method.

[0125] In some embodiments, an embodiment of the present invention also provides an electronic device, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the intelligent monitoring method for orthopedic rehabilitation.

[0126] Figure 7 FIG. 1 is a schematic diagram of the hardware structure of an electronic device for executing an intelligent monitoring method for orthopedic rehabilitation provided by another embodiment of the present invention. Figure 7 As shown, the device includes:

[0127] One or more processors 710 and memory 720, Figure 7 A processor 710 is taken as an example.

[0128] The device for executing the orthopedic rehabilitation intelligent monitoring method may further include: an input device 730 and an output device 740 .

[0129] The processor 710, the memory 720, the input device 730 and the output device 740 may be connected via a bus or other means. Figure 7 The example of connecting through bus is taken in the following.

[0130] The memory 720 is a non-volatile computer-readable storage medium that can be used to store non-volatile software programs, non-volatile computer executable programs and modules, such as program instructions / modules corresponding to the orthopedic rehabilitation intelligent monitoring method in the embodiment of the present invention. The processor 710 executes various functional applications and data processing of the server by running the non-volatile software programs, instructions and modules stored in the memory 720, that is, the orthopedic rehabilitation intelligent monitoring method of the above method embodiment is implemented.

[0131] The memory 720 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the electronic device, etc. In addition, the memory 720 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the memory 720 may optionally include a memory remotely arranged relative to the processor 710, and these remote memories may be connected to the electronic device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0132] The input device 730 may receive input digital or character information and generate signals related to user settings and function control of the electronic device. The output device 740 may include a display device such as a display screen.

[0133] The one or more modules are stored in the memory 720, and when executed by the one or more processors 710, the orthopedic rehabilitation intelligent monitoring method in any of the above method embodiments is executed.

[0134] The above product can execute the method provided by the embodiment of the present invention, and has the functional modules and beneficial effects corresponding to the execution method. For technical details not described in detail in this embodiment, please refer to the method provided by the embodiment of the present invention.

[0135] The electronic devices of the embodiments of the present invention exist in various forms, including but not limited to:

[0136] (1) Mobile communication equipment: This type of equipment is characterized by having mobile communication functions and its main purpose is to provide voice and data communications. This type of terminal includes: smart phones, multimedia phones, functional phones, and low-end phones.

[0137] (2) Ultra-mobile personal computer devices: These devices belong to the category of personal computers, have computing and processing functions, and generally also have mobile Internet access features. These terminals include: PDA, MID and UMPC devices, etc.

[0138] (3) Portable entertainment devices: These devices can display and play multimedia content. They include audio and video players, handheld game consoles, e-books, smart toys, and portable car navigation devices.

[0139] (4) Other onboard electronic devices with data interaction functions, such as on-board devices installed in vehicles.

[0140] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0141] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a general hardware platform, and of course, by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the relevant technology can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0142] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An intelligent monitoring method for orthopedic rehabilitation, characterized in that: The method comprises: Acquire user attribute data and user rehabilitation plan of the target user to be monitored; the user rehabilitation plan includes a plurality of consecutive planned time periods and corresponding simulated posture simulation models; each of the planned time periods has a corresponding rehabilitation action, and the simulated posture simulation model is a simulation model that simulates the target user when performing the corresponding rehabilitation action; For each simulated posture simulation model, the user attribute data, the simulated posture simulation model and the rehabilitation action are input into a preset device stimulation evaluation model to determine the predicted stress stimulation data of the orthopedic rehabilitation device corresponding to the simulated posture simulation model; the predicted stress stimulation data represents the predicted stress stimulation data that the target user will generate on the orthopedic rehabilitation device by performing the rehabilitation action; In the case where it is detected that the orthopedic rehabilitation device samples measured stress stimulation data, matching a target planned time period from the multiple planned time periods according to the sampling time of the measured stress stimulation data, and identifying whether the target user has an abnormal rehabilitation state according to the predicted stress stimulation data corresponding to the target planned time period and the measured stress stimulation data; When an abnormal rehabilitation state is identified, an early warning notification is sent to a preset medical terminal according to the measured stress stimulation data of the target user.

2. The method according to claim 1, characterized in that The step of identifying whether the target user is in an abnormal rehabilitation state according to the predicted stress stimulation data corresponding to the target planned time period and the measured stress stimulation data includes: Calculating a stress deviation value between the predicted stress stimulation data and the measured stress stimulation data; When it is detected that the stress deviation value exceeds a preset stress deviation threshold, it is determined that the target user is in an abnormal rehabilitation state.

3. The method according to claim 1, characterized in that: The device stimulation evaluation model adopts a multimodal fusion prediction model based on ensemble learning and self-attention mechanism, and the multimodal fusion prediction model includes an input layer, a base learning module, a feature fusion layer and an output layer; Wherein, the base learning module comprises a random forest layer, a bidirectional long short-term memory network layer and a convolutional neural network layer; The input layer is used to receive the user attribute data, the simulated posture simulation model and the rehabilitation action; The random forest layer is used to perform integrated prediction processing on the user attribute data through multiple decision trees to extract static user attribute feature representation Y RF ; The bidirectional long short-term memory network layer is used to capture the time dependency relationship corresponding to the rehabilitation action to obtain the corresponding time series feature representation Y Bi-LSTM : Forward LSTM processing: Backward LSTM processing: Merge the hidden state of the forward LSTM and the hidden state of the backward LSTM to form the output feature representation of the bidirectional LSTM: In the formula, the rehabilitation action is obtained through the time series data S = [s1, s2, ..., s T ] is defined as follows, T represents the total number of time steps included in the rehabilitation action; s t is the rehabilitation action feature vector at the tth time step; represents the hidden state of the forward LSTM at the tth time step, represents the hidden state of the forward LSTM at the t-1th time step; represents the hidden state of the backward LSTM at the tth time step, represents the hidden state of the backward LSTM at the t+1th time step.

4. The method according to claim 3, characterized in that: The convolutional neural network layer is used to extract the model space features corresponding to the simulated posture simulation model through the residual network structure: The residual network structure includes multiple residual blocks, each of which is implemented by a skip connection: Y res =F(X res )+X res Where Y res represents the output of the residual block, X res Represents the input of the residual block, and F represents the combined operation of convolution, batch normalization, and activation function; The feature fusion layer is used to perform weighted fusion on the output features of the base learning module to obtain the corresponding deep fusion features: F =α1Y CNN +α2Y Bi-LSTM +α3Y RF In the formula, F represents the deep fusion feature, Y CNN Represents the model space features of the final output of the convolutional neural network layer; α1, α2, and α3 represent adaptive fusion weights, which can be dynamically adjusted through the back-propagation algorithm to ensure that the importance of different features is reasonably reflected in the fusion process.

5. The method according to claim 4, characterized in that: The output layer is used to perform nonlinear transformation on the deep fusion features through a fully connected layer to predict and output stress stimulation data of the simulated posture simulation model: Z=σ(W1F+b1) Y pred =σ(W2Z+b2) In the formula, σ represents the ReLU activation function, Y pred represents the predicted stress stimulation data of the simulated posture simulation model; W1, b1 respectively represent the weight matrix and bias term of the first fully connected layer, W2, b2 respectively represent the weight matrix and bias term of the second fully connected layer; Z represents the feature representation of the first fully connected layer after nonlinear transformation through the first fully connected layer.

6. The method according to claim 1 or 3, characterized in that: The loss function L of the device stimulus evaluation model is 综合 for: L 综合 =λ1L DTW +λ2L MSE Where, L DTW and L MSE They represent dynamic time warping loss and mean square error loss respectively, λ1 and λ2 represent the weight parameters of the corresponding loss terms respectively; N represents the total number of samples in the data sample set, Y pred,i represents the predicted stress stimulation data of the i-th sample in the data sample set, Y true,i represents the true stress stimulation data of the i-th sample in the data sample set, Y true,j represents the real stress stimulation data of the jth sample in the data sample set; δ(i,j) is the path constraint condition, indicating whether the path corresponding to data sample i and data sample j is allowed at the corresponding time step; d(Y pred,i ,Y true,j ) indicates Y pred,i and Y true,j The Euclidean distance between .

7. The method according to claim 1, characterized in that The step of obtaining user attribute data and user rehabilitation plan of the target user to be monitored includes: Acquire input information, and parse the user attribute data and user rehabilitation plan of the target user to be monitored corresponding to the input information; the user rehabilitation plan includes a plurality of consecutive planned time periods and corresponding planned rehabilitation action names; Determining target standard posture simulation models that match the names of the planned rehabilitation movements from a posture model library; the posture model library includes a plurality of rehabilitation movement names and corresponding standard posture simulation models; For each of the target standard posture simulation models, the target standard posture simulation model and the user attribute data are input into a posture conversion network to convert and generate a corresponding simulated posture simulation model; The user rehabilitation program is generated according to a plurality of consecutive planned time periods and the corresponding simulated posture simulation model.

8. The method according to claim 7, characterized in that The posture conversion network adopts a conditional generative adversarial network, which includes an input layer, a posture encoder, an attribute encoder, a feature fusion layer, a generator and a discriminator; The input layer is used to receive the target standard posture simulation model P std and user attribute data U; The posture encoder is used to extract the standard posture coding features corresponding to the target standard posture: F p1 =σ(Conv(P std )) F p2 =Pool(F p1 ) F p =σ(W p ·F p2 +b p ) In the formula, Conv represents the convolution operation, σ represents the ReLU activation function, and F p1 represents the posture convolution feature; Pool represents the pooling operation, F p2 Represents the features of the pooling layer output; W p and b p Denote the weight matrix and bias term of the fully connected layer of the pose encoding, F p Represents standard posture coding features; The attribute encoder is used to extract the user attribute encoding features corresponding to the user attribute data: F u =σ(W u ·U+b u ) Where W u and b u Represent the weight matrix and bias term of the attribute encoding fully connected layer, F u Represents the user attribute encoding feature.

9. The method according to claim 8, characterized in that The feature fusion layer is used to adaptively fuse the standard posture coding features and the user attribute coding features to determine corresponding multimodal fusion features: F fused =γ·F p +(1-γ)·F u In the formula, γ represents the adaptive fusion weight, F fused Represents multimodal fusion features; The generator is used to generate a personalized simulated posture simulation model according to the multimodal fusion features: F g1 =σ(W g1 ·F fused +b g1 ) P sim =Deconv(F g1 ) Where W g1 and b g1 Represent the weight matrix and bias term of the fully connected layer of the generator respectively; F g1 represents the high-dimensional representation of the generator; Deconv represents the deconvolution operation, P sim Represents the simulated posture simulation model.

Images