Orthopedic patient rehabilitation progress tracking method and system

Abstract

The invention relates to the technical field of orthopedic health monitoring, and provides an orthopedic patient rehabilitation progress tracking method and system. The system comprises a data acquisition module, a central processing unit module and a visual report generation module, the data acquisition module is used for acquiring physiological signal data, image data and subjective feedback data of a patient; the central processing unit module performs preprocessing and feature extraction on the data, and performs rehabilitation progress prediction by using an improved BiTCN-GRU model; in addition, an emotion-driven mechanism is adopted to optimize a position updating rule of a cat group optimization algorithm, and hyper-parameters of the BiTCN-GRU model are optimized; the visual report generation module generates rehabilitation suggestions based on the optimized prediction result; according to the method and the system, the rehabilitation progress prediction precision can be remarkably improved, a personalized rehabilitation scheme is provided, and the rehabilitation process of the orthopedic patient is optimized.

Description

Technical Field

[0001] The present invention relates to the technical field of orthopedic health monitoring, and particularly to a method and system for tracking the rehabilitation progress of orthopedic patients. Background Art

[0002] With the continuous development of technology, more and more intelligent technologies and data analysis methods have been applied to the tracking of the rehabilitation progress of orthopedic patients, aiming to improve the accuracy of assessment and rehabilitation effects. However, traditional orthopedic rehabilitation progress tracking systems still have significant deficiencies. Traditional systems usually rely on a single data source for rehabilitation assessment. Although these systems can provide certain rehabilitation progress feedback, due to the lack of comprehensive analysis and processing of multi-modal data, their performance in terms of accuracy, sensitivity, and personalization is limited. First, traditional methods often ignore the effective combination of multi-dimensional data of patients, such as physiological signal data, imaging data, and subjective feedback data, resulting in insufficient accuracy in predicting the rehabilitation progress and being unable to fully reflect the rehabilitation status of patients. Second, traditional methods usually adopt simple feature extraction and prediction models, which fail to effectively capture the complex time-dependence and individual differences in the patient's rehabilitation process, resulting in generally rough prediction results of the rehabilitation progress and being difficult to provide precise rehabilitation guidance. Therefore, there is an urgent need for an improved orthopedic patient rehabilitation progress tracking system that can address the deficiencies in multi-modal data fusion, noise processing, and feature extraction in traditional systems and improve the accuracy and real-time performance of rehabilitation progress prediction. Summary of the Invention

[0003] The present invention provides an orthopedic patient rehabilitation progress tracking system, which includes a data acquisition module, a central processing unit module, and a visualization report generation module. The data acquisition module is used to collect physiological signal data, subjective feedback data, and imaging data of orthopedic patients and integrate these multi-modal data. The central processing unit module preprocesses and extracts features from the integrated data and uses an improved BiTCN-GRU model for rehabilitation progress prediction. The visualization report generation module generates a rehabilitation progress curve, a joint range of motion report, and an imaging change report based on the prediction results, providing intuitive rehabilitation feedback and personalized rehabilitation suggestions for patients and doctors.

[0004] The present invention also provides a method for tracking the rehabilitation progress of orthopedic patients; after data preprocessing, the system optimizes the forward local feature extraction of the BiTCN model through the dynamic routing mechanism of the capsule network, enhancing the system's ability to capture detailed features; at the same time, a dilated convolutional pyramid is used to replace the dilated causal convolution to optimize the reverse local feature extraction of the BiTCN model, improving the system's ability to model multi-scale temporal dependencies, thereby improving the prediction accuracy; in addition, the present invention introduces an emotion-driven mechanism to optimize the position update rule of the cat swarm optimization algorithm, significantly improving the prediction accuracy of the system; through these technical improvements, the present invention can significantly improve the accuracy of rehabilitation progress prediction and provide a more personalized and refined rehabilitation plan.

[0005] A method for tracking the rehabilitation progress of orthopedic patients provided by the present invention includes the following steps:

[0006] Step S1: Data collection: Collect the physiological signal data, subjective feedback data, and image data of orthopedic patients through a data collection module, and integrate them to obtain multi-modal patient rehabilitation data; the physiological signal data includes electromyogram signal data, joint range of motion data, and gait analysis data, the subjective feedback data includes pain score and function score, and the image data includes X-ray image data, MRI image data, and CT image data;

[0007] Step S2: Data preprocessing: In the central processing unit module, denoise, image enhancement, and standardization are performed on the multi-modal patient rehabilitation data to generate preprocessed multi-modal patient rehabilitation data, which includes preprocessed physiological signal data, preprocessed subjective feedback data, and preprocessed image data;

[0008] Step S3: Feature extraction: In the central processing unit module, a ResNet model is established as a pre-trained model, and the ResNet model is used to extract features from the preprocessed image data to generate image feature data; extract the time-domain features, frequency-domain features, and dynamic features of the preprocessed physiological signal data to generate physiological signal feature data; integrate the image feature data, physiological signal feature data, and preprocessed subjective feedback data as multi-modal patient rehabilitation feature data;

[0009] Step S4: Rehabilitation progress prediction: In the central processing unit module, a BiTCN-GRU model is established. The forward local feature extraction of the BiTCN model is optimized through the dynamic routing mechanism of the capsule network, and the reverse local feature extraction of the BiTCN model is optimized by replacing the dilated causal convolution with a dilated convolutional pyramid to construct an enhanced BiTCN model. The hyperparameters of the enhanced BiTCN model are initialized, and the enhanced BiTCN model replaces the BiTCN model in the BiTCN-GRU model to construct an improved BiTCN-GRU model. The multi-modal patient rehabilitation feature data is input into the improved BiTCN-GRU model, and the rehabilitation progress prediction result is output;

[0010] Step S5: Model optimization: In the central processing unit module, an emotion-driven mechanism is introduced to optimize the position update rule of the cat swarm optimization algorithm to construct an emotion-cat swarm optimization algorithm. The hyperparameters of the improved BiTCN-GRU model are optimized according to the emotion-cat swarm optimization algorithm to generate the optimal hyperparameter configuration, and the rehabilitation progress prediction result is improved according to the optimal hyperparameter configuration to generate the optimized rehabilitation progress prediction result;

[0011] Step S6: Visual report generation: In the visual report generation module, rehabilitation progress prediction and personalized rehabilitation suggestions are made according to the optimized rehabilitation progress prediction result, and a rehabilitation progress curve, a joint range of motion report, and an imaging change report are generated.

[0012] Further, the improved BiTCN-GRU model in step S4 includes an enhanced BiTCN model, a GRU layer, a Dropout layer, and a fully connected layer, and a rehabilitation progress prediction space is defined. The rehabilitation progress prediction space includes the time required for rehabilitation, the rehabilitation progress level, the joint range of motion recovery value, the muscle strength recovery value, and the function recovery index.

[0013] Further, step S4 specifically includes the following steps:

[0014] Step S41: Extract the local features of the multi-modal patient rehabilitation feature data through the enhanced BiTCN model to obtain comprehensive time series feature data;

[0015] Step S42: Perform time series analysis on the comprehensive time series feature data through the GRU layer; capture the long-term and short-term time dependencies of the comprehensive time series feature data through causality to obtain time-dependent feature data;

[0016] Step S43: Process the time-dependent feature data through the Dropout layer, and perform regularization processing by randomly discarding some neurons to prevent overfitting to obtain regularized time series feature data;

[0017] Step S44: Map the regularized time series feature data to the rehabilitation progress prediction space through a fully connected layer, complete the output conversion from features to rehabilitation progress results, and obtain the rehabilitation progress prediction results.

[0018] Further, step S41 specifically includes the following steps:

[0019] Step S411: Data organization: Organize the multi-modal patient rehabilitation feature data into multi-dimensional time series feature data in chronological order;

[0020] Step S412: Forward local feature extraction: Process the multi-dimensional time series feature data through dilated causal convolution to expand the receptive field, extract local dependence features in the time dimension, obtain dilated convolution feature data, represent the dilated convolution feature data as vectorized capsule data, and capture the spatio-temporal relationship of the vectorized capsule data through the dynamic routing mechanism of the capsule network to generate the forward convolution result. The formula used is as follows:

[0021] ;

[0022] Among them, represents the forward convolution result, represents the activation function, represents the number of capsules, represents the capsule index, represents the th feature vector of the capsule, represents the th feature vector after weighted aggregation of the capsule through the dynamic routing mechanism, represents the L2 norm of, represents the scaling factor, represents the normalized vector;

[0023] Step S413: Forward feature processing: Perform batch normalization, weight normalization, non-linear mapping, and regularization processing on the forward convolution result, extract the historical dependence relationship in the multi-dimensional time series feature data, and generate forward time dependence feature data;

[0024] Step S414: Reverse local feature extraction: Process the multi-dimensional time series feature data through a dilated convolutional pyramid, use dilated convolutions with different dilation rates to form a pyramid structure, obtain convolution results with different dilation rates, further expand the receptive field, capture the multi-scale time dependence relationship of the multi-dimensional time series feature data, and fuse the convolution results with different dilation rates to generate the reverse convolution result. The formula used is as follows:

[0025] Dilated convolution calculation formula:

[0026] ;

[0027] Among them, represents the time step index, represents the dilation rate index of the dilated convolution, represents the convolutional kernel index, represents the convolutional kernel size, represents the dilation rate of the th layer of convolution, represents the feature result after the dilated convolution, represents the feature at the th time step in the multi-dimensional time series feature data;

[0028] Step S414: Reverse feature processing: Perform batch normalization, weight normalization, non-linear mapping, and regularization on the reverse convolution result, extract the future dependencies in the multi-dimensional time series feature data, and generate reverse time-dependent feature data;

[0029] Step S414: Feature fusion: Fuse the forward time-dependent feature data and the reverse time-dependent feature data to generate bidirectional time-dependent feature data;

[0030] Step S415: Feature adjustment: Adjust the dimensions of the bidirectional time-dependent feature data to convert it into comprehensive time series feature data.

[0031] Furthermore, Step S5 specifically includes the following steps:

[0032] Step S51: Initialize the population: Randomly generate cat individuals in the population of the cat swarm optimization algorithm. Each cat individual represents a set of hyperparameter configurations of the improved BiTCN-GRU model, define the search space, and generate the initial cat population according to the search space;

[0033] Step S52: Set the fitness function: Define the fitness function, calculate the initial fitness values of the initial cat population through the fitness function, and select the current global optimal solution and the local best solution according to the initial fitness values;

[0034] Step S53: Position update: Introduce an emotion-driven mechanism to simulate the emotional fluctuations of cat individuals during the search process, optimize the position update rule of the cat swarm optimization algorithm, construct an emotion-driven - position update rule, adjust the position of each cat individual according to the emotion-driven - position update rule, and the cat individual updates its position according to the current global optimal solution and the local best solution to generate the updated cat population. The formula used is as follows:

[0035] ;

[0036] Among them, represents the cat individual index, Represents a cat individual The velocity updated in the current iteration Represents a cat individual The current velocity Represents the inertia weight Represents the individual learning factor and Represents a random number Represents a cat individual The local best position of the cat individual, i.e., the local best solution Represents the swarm learning factor Represents the current global best position, i.e., the current global optimal solution Represents the emotion-driven velocity change, which adjusts the velocity according to the emotional state of the cat individual;

[0037] Step S54: Genetic operation: Perform crossover operation and mutation operation on the updated cat population to increase the diversity of the population and generate a genetic cat population; Calculate the genetic fitness value of the genetic cat population according to the fitness function, and optimize the current global optimal solution and local best solution according to the genetic fitness value;

[0038] Step S55: Iterative optimization: Set the maximum number of iterations, and repeat Steps S53 to S54 until the maximum number of iterations is reached to generate the global optimal solution, obtain the optimal hyperparameter configuration, improve the rehabilitation progress prediction result according to the optimal hyperparameter configuration, and generate the optimized rehabilitation progress prediction result.

[0039] The present invention provides an orthopedic patient rehabilitation progress tracking system for implementing the above method. The system includes a data acquisition module, a central processing unit module, and a visualization report generation module; The data acquisition module collects the patient's rehabilitation data and transmits it to the central processing unit module. The central processing unit module extracts features and makes predictions on the patient's rehabilitation data, and transmits the prediction result to the visualization report generation module. The visualization report generation module generates a rehabilitation progress curve, a joint range of motion report, and an imaging change report according to the prediction result.

[0040] Adopting the above solution, the beneficial effects obtained by the present invention are as follows:

[0041] The present invention optimizes the forward local feature extraction of the BiTCN model by introducing the dynamic routing mechanism of the capsule network, solving the problem in traditional systems of difficultly and precisely capturing the subtle changes in the rehabilitation progress of orthopedic patients; the dynamic routing mechanism can effectively enhance the system's modeling ability for complex features, especially when dealing with multi-modal data, it can more finely extract local features in images, signals, and subjective feedback data, improving the accuracy of rehabilitation progress prediction; by combining the dilated convolutional pyramid to replace the dilated causal convolution, the backward local feature extraction of the BiTCN model is further optimized, making the system more sensitive in capturing multi-scale time-dependent relationships and being able to more comprehensively analyze the dynamic changes during the rehabilitation process, significantly improving the performance of the present invention in predicting the rehabilitation progress of orthopedic patients; by improving the BiTCN-GRU model, the present invention enhances the prediction accuracy of the rehabilitation progress of orthopedic patients; the system makes full use of the relationships between multi-modal data and demonstrates superior performance in feature extraction and time series analysis; this improvement not only enhances the system's processing ability for complex data but also effectively improves the real-time response ability in rehabilitation progress prediction, enabling the system to more accurately reflect the patient's rehabilitation situation, thus helping to develop more precise personalized rehabilitation plans;

[0042] In addition, the present invention introduces an emotion-driven mechanism to optimize the position update rule of the cat swarm optimization algorithm, further enhancing the optimization effect of the hyperparameters of the BiTCN-GRU model of the system; by simulating the emotional fluctuations of cat individuals to adjust the position update, it not only improves the search efficiency of the cat swarm optimization algorithm but also solves the problem of being trapped in local optima in hyperparameter optimization; this mechanism can ensure finding the optimal hyperparameter configuration in multiple iterations, improving the accuracy of the rehabilitation progress prediction results in the system; through these technical improvements, the present invention has made significant breakthroughs in the accuracy of predicting the rehabilitation progress of orthopedic patients, the formulation of personalized recommendations, and the real-time monitoring ability, providing a more scientific and personalized rehabilitation management plan for orthopedic patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a schematic diagram of the model of an orthopedic patient rehabilitation progress tracking system proposed by the present invention;

[0044] Figure 2 It is a schematic diagram of the process of an orthopedic patient rehabilitation progress tracking method proposed by the present invention;

[0045] Figure 3 It is a schematic diagram of the structure of the enhanced BiTCN model in Embodiment 6;

[0046] Figure 4 It is a schematic diagram of the process of the emotion-cat swarm optimization algorithm in Embodiment 8. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0047] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0048] Embodiment 1. According to Figure 1 , the present invention provides an orthopedic patient rehabilitation progress tracking system, which includes a data acquisition module, a central processor module, and a visualization report generation module. The data acquisition module is used to collect physiological signal data, subjective feedback data, and imaging data of orthopedic patients, and integrate these multi-modal data. The central processor module preprocesses and extracts features from the integrated data, and uses an improved BiTCN-GRU model for rehabilitation progress prediction. The visualization report generation module generates a rehabilitation progress curve, a joint range of motion report, and an imaging change report according to the prediction results, providing intuitive rehabilitation feedback and personalized rehabilitation suggestions for patients and doctors.

[0049] Embodiment 2. According to Figure 2 , a method for tracking the rehabilitation progress of orthopedic patients provided by the present invention includes the following steps:

[0050] Step S1: Data acquisition: Collect physiological signal data, subjective feedback data, and image data of orthopedic patients through the data acquisition module, and integrate them to obtain multi-modal patient rehabilitation data. The physiological signal data includes electromyogram signal data, joint range of motion data, and gait analysis data. The subjective feedback data includes pain score and function score. The image data includes X-ray image data, MRI image data, and CT image data.

[0051] Step S2: Data preprocessing: In the central processor module, denoise, image enhancement, and standardization are performed on the multi-modal patient rehabilitation data to generate preprocessed multi-modal patient rehabilitation data, which includes preprocessed physiological signal data, preprocessed subjective feedback data, and preprocessed image data.

[0052] Step S3: Feature extraction: In the central processor module, a ResNet model is established as a pre-trained model, and the ResNet model is used to extract features from the preprocessed image data to generate image feature data. Extract the time-domain features, frequency-domain features, and dynamic features of the preprocessed physiological signal data to generate physiological signal feature data. Integrate the image feature data, physiological signal feature data, and preprocessed subjective feedback data as multi-modal patient rehabilitation feature data.

[0053] Step S4: Rehabilitation progress prediction: In the central processing unit module, a BiTCN-GRU model is established. The forward local feature extraction of the BiTCN model is optimized through the dynamic routing mechanism of the capsule network, and the reverse local feature extraction of the BiTCN model is optimized by replacing the dilated causal convolution with a dilated convolutional pyramid to construct an enhanced BiTCN model. The hyperparameters of the enhanced BiTCN model are initialized, and the enhanced BiTCN model replaces the BiTCN model in the BiTCN-GRU model to construct an improved BiTCN-GRU model. The multi-modal patient rehabilitation feature data is input into the improved BiTCN-GRU model, and the rehabilitation progress prediction result is output;

[0054] Step S5: Model optimization: In the central processing unit module, an emotion-driven mechanism is introduced to optimize the position update rule of the cat swarm optimization algorithm to construct an emotion-cat swarm optimization algorithm. The hyperparameters of the improved BiTCN-GRU model are optimized according to the emotion-cat swarm optimization algorithm to generate the optimal hyperparameter configuration, and the rehabilitation progress prediction result is improved according to the optimal hyperparameter configuration to generate the optimized rehabilitation progress prediction result;

[0055] Step S6: Visual report generation: In the visual report generation module, rehabilitation progress prediction and personalized rehabilitation suggestions are made according to the optimized rehabilitation progress prediction result, and a rehabilitation progress curve, a joint range of motion report, and an imaging change report are generated.

[0056] Embodiment 3: This embodiment is based on Embodiment 2. In this embodiment, the improved BiTCN-GRU model in Step S4 includes an enhanced BiTCN model, a GRU layer, a Dropout layer, and a fully connected layer. A rehabilitation progress prediction space is defined, and the rehabilitation progress prediction space includes the time required for rehabilitation, the rehabilitation progress level, the joint range of motion recovery value, the muscle strength recovery value, and the function recovery index.

[0057] Embodiment 4: This embodiment is based on Embodiment 3. In this embodiment, Step S4 specifically includes the following steps:

[0058] Step S41: The local features of the multi-modal patient rehabilitation feature data are extracted through the enhanced BiTCN model to obtain comprehensive time series feature data;

[0059] Step S42: Time series analysis is performed on the comprehensive time series feature data through the GRU layer; the long-term and short-term time dependencies of the comprehensive time series feature data are captured through causal relationships to obtain time-dependent feature data;

[0060] Step S43: The time-dependent feature data is processed through the Dropout layer, and regularization processing is performed by randomly discarding some neurons to prevent overfitting, and regularized time series feature data is obtained;

[0061] Step S44: Map the regularized time series feature data to the rehabilitation progress prediction space through a fully connected layer, complete the output conversion from features to rehabilitation progress results, and obtain the rehabilitation progress prediction result.

[0062] Example 5. This example is based on Example 3. In this example, step S4 specifically includes the following steps:

[0063] Step S41: Extract the local features of the multi-modal patient rehabilitation feature data through the BiTCN model to obtain the comprehensive time series feature data;

[0064] Step S42: Perform time series analysis on the comprehensive time series feature data through the GRU layer; capture the long-term and short-term time dependencies of the comprehensive time series feature data through causal relationships to obtain the time-dependent feature data;

[0065] Step S43: Process the time-dependent feature data through the Dropout layer, perform regularization processing by randomly discarding some neurons to prevent overfitting, and obtain the regularized time series feature data;

[0066] Step S44: Map the regularized time series feature data to the rehabilitation progress prediction space through a fully connected layer, complete the output conversion from features to rehabilitation progress results, and obtain the rehabilitation progress prediction result.

[0067] Example 6. According to Figure 3 , this example is based on Example 4. In this example, step S41 specifically includes the following steps:

[0068] Step S411: Data organization: Organize the multi-modal patient rehabilitation feature data into multi-dimensional time series feature data in chronological order;

[0069] Step S412: Forward local feature extraction: Process the multi-dimensional time series feature data through dilated causal convolution to expand the receptive field, extract the local dependence features in the time dimension to obtain the dilated convolution feature data, represent the dilated convolution feature data as vectorized capsule data, and capture the spatio-temporal relationship of the vectorized capsule data through the dynamic routing mechanism of the capsule network to generate the forward convolution result. The formula used is as follows:

[0070] ;

[0071] Among them, represents the forward convolution result, represents the activation function, represents the number of capsules, represents the capsule index, represents the th capsule's feature vector. Denote the feature vector after weighted aggregation of the th capsule through the dynamic routing mechanism, denote the L2 norm of, denote the scaling factor, denote the normalized vector;

[0072] Step S413: Forward feature processing: Perform batch normalization, weight normalization, non-linear mapping, and regularization on the forward convolution result, extract the historical dependencies in the multi-dimensional time series feature data, and generate forward time-dependent feature data;

[0073] Step S414: Backward local feature extraction: Process the multi-dimensional time series feature data through the dilated convolution pyramid, use dilated convolutions with different dilation rates to form a pyramid structure, obtain the convolution results with different dilation rates, further expand the receptive field, capture the multi-scale time dependencies of the multi-dimensional time series feature data, and fuse the convolution results with different dilation rates to generate the backward convolution result. The formula used is as follows:

[0074] Dilated convolution calculation formula:

[0075] ;

[0076] where, denotes the time step index, denotes the dilation rate index of the dilated convolution, denotes the convolution kernel index, denotes the convolution kernel size, denotes the dilation rate of the th layer of convolution, denotes the feature result after dilated convolution, denotes the feature at the th time step in the multi-dimensional time series feature data, denotes the convolution kernel weight;

[0077] Step S414: Backward feature processing: Perform batch normalization, weight normalization, non-linear mapping, and regularization on the backward convolution result, extract the future dependencies in the multi-dimensional time series feature data, and generate backward time-dependent feature data;

[0078] Step S414: Feature fusion: Fuse the forward time-dependent feature data and the backward time-dependent feature data to generate two-way time-dependent feature data;

[0079] Step S415: Feature adjustment: Adjust the dimensions of the two-way time-dependent feature data to convert it into comprehensive time series feature data.

[0080] Example 7. This example is based on Example 4. In this example, step S41 specifically includes the following steps:

[0081] Step S411: Data organization: Organize the multi-modal patient rehabilitation feature data into multi-dimensional time series feature data in chronological order;

[0082] Step S412: Forward local feature extraction: Process the multi-dimensional time series feature data through dilated causal convolution to expand the receptive field, extract the local dependence features in the time dimension, obtain the dilated convolution feature data, represent the dilated convolution feature data as vectorized capsule data, and capture the spatio-temporal relationship of the vectorized capsule data through the dynamic routing mechanism of the capsule network to generate the forward convolution result. The formula used is as follows:

[0083] ;

[0084] Among them, represents the forward convolution result, represents the activation function, represents the number of capsules, represents the capsule index, represents the th feature vector of the capsule, represents the feature vector after weighted aggregation of the th capsule through the dynamic routing mechanism, represents the L2 norm of, represents the scaling factor, represents the normalized vector;

[0085] Step S413: Forward feature processing: Perform batch normalization, weight normalization, non-linear mapping, and regularization on the forward convolution result to extract the historical dependence relationship in the multi-dimensional time series feature data and generate the forward time dependence feature data;

[0086] Step S414: Reverse local feature extraction: Process the multi-dimensional time series feature data through dilated causal convolution to generate the reverse convolution result;

[0087] Step S414: Reverse feature processing: Perform batch normalization, weight normalization, non-linear mapping, and regularization on the reverse convolution result to extract the future dependence relationship in the multi-dimensional time series feature data and generate the reverse time dependence feature data;

[0088] Step S414: Feature fusion: Fusion the forward time dependence feature data and the reverse time dependence feature data to generate the bidirectional time dependence feature data;

[0089] Step S415: Feature adjustment: Adjust the dimensions of the two-way time-dependent feature data and convert it into comprehensive time series feature data.

[0090] Example 8. According to Figure 4 , this example is based on Example 6. In this example, step S5 specifically includes the following steps:

[0091] Step S51: Initialize the population: Randomly generate cat individuals in the population of the cat swarm optimization algorithm. Each cat individual represents a set of hyperparameter configurations of the improved BiTCN-GRU model. Define the search space and generate the initial cat population according to the search space;

[0092] Step S52: Set the fitness function: Define the fitness function, calculate the initial fitness values of the initial cat population through the fitness function, and select the current global optimal solution and the local best solution according to the initial fitness values;

[0093] Step S53: Position update: Introduce an emotion-driven mechanism to simulate the emotional fluctuations of cat individuals during the search process, optimize the position update rule of the cat swarm optimization algorithm, construct an emotion-driven - position update rule, and adjust the position of each cat individual according to the emotion-driven - position update rule. The cat individual updates its position according to the current global optimal solution and the local best solution to generate the updated cat population. The formula used is as follows:

[0094] ;

[0095] Where represents the cat individual index, represents the cat individual 's updated speed in the current iteration, represents the cat individual 's current speed, represents the inertia weight, represents the individual learning factor, and represent random numbers, represents the local best position of the cat individual , that is, the local best solution, represents the swarm learning factor, represents the current global best position, that is, the current global optimal solution, represents the speed change driven by emotion, and adjusts the speed according to the emotional state of the cat individual;

[0096] Step S54: Genetic operation: Perform crossover operation and mutation operation on the updated cat population to increase the diversity of the population and generate a genetic cat population; Calculate the genetic fitness values of the genetic cat population according to the fitness function, and optimize the current global optimal solution and the local best solution according to the genetic fitness values;

[0097] Step S55: Iterative optimization: Set the maximum number of iterations, and repeatedly execute Step S53 to Step S54 until the maximum number of iterations is reached, generate the global optimal solution, obtain the optimal hyperparameter configuration, improve the rehabilitation progress prediction result according to the optimal hyperparameter configuration, and generate the optimized rehabilitation progress prediction result.

[0098] Embodiment Nine. This embodiment is based on Embodiment Six. In this embodiment, Step S5 specifically includes the following steps:

[0099] Step S51: Initialize the population: Randomly generate cat individuals in the population of the cat swarm optimization algorithm. Each cat individual represents a set of hyperparameter configurations of the improved BiTCN-GRU model. Define the search space and generate the initial cat population according to the search space;

[0100] Step S52: Set the fitness function: Define the fitness function, calculate the initial fitness values of the initial cat population through the fitness function, and select the current global optimal solution and local best solution according to the initial fitness values;

[0101] Step S53: Position update: According to the position update rule of the optimized cat swarm optimization algorithm, adjust the position of each cat individual. The cat individual updates its position according to the current global optimal solution and local best solution, and generates the updated cat population;

[0102] Step S54: Genetic operation: Perform crossover operation and mutation operation on the updated cat population to increase the diversity of the population, and generate the genetic cat population; Calculate the genetic fitness values of the genetic cat population according to the fitness function, and optimize the current global optimal solution and local best solution according to the genetic fitness values;

[0103] Step S55: Iterative optimization: Set the maximum number of iterations, and repeatedly execute Step S53 to Step S54 until the maximum number of iterations is reached, generate the global optimal solution, obtain the optimal hyperparameter configuration, improve the rehabilitation progress prediction result according to the optimal hyperparameter configuration, and generate the optimized rehabilitation progress prediction result.

[0104] Embodiment Nine. This embodiment is based on the above-mentioned embodiment. In this embodiment, the present invention provides an orthopedic patient rehabilitation progress tracking system for implementing the above method. The system includes a data acquisition module, a central processor module, and a visualization report generation module; the data acquisition module acquires the patient's rehabilitation data and transmits it to the central processor module. The central processor module extracts features and makes predictions on the patient's rehabilitation data, and transmits the prediction results to the visualization report generation module. The visualization report generation module generates a rehabilitation progress curve, a joint range of motion report, and an imaging change report according to the prediction results.

[0105] The above description of the present invention and its implementation manners is not restrictive. What is shown in the drawings is only one of the implementation manners of the present invention, and the actual structure is not limited thereto. Generally speaking, if those of ordinary skill in the art are inspired by it and design, without creative efforts, structural manners and embodiments similar to the technical solution without departing from the gist of the present invention, they shall fall within the protection scope of the present invention.

Claims

1. A method for tracking the rehabilitation progress of orthopedic patients, characterized by: The method specifically comprises the following steps: Step S1: Data collection: collecting multimodal patient rehabilitation data; Step S2: data preprocessing: preprocessing the multimodal patient rehabilitation data to generate preprocessed multimodal patient rehabilitation data; Step S3: feature extraction: extracting features from the preprocessed multimodal patient rehabilitation data to generate multimodal patient rehabilitation feature data; Step S4: Rehabilitation progress prediction: In the central processing unit module, a BiTCN-GRU model is established, the forward local feature extraction of the BiTCN model is optimized through the dynamic routing mechanism of the capsule network, the reverse local feature extraction of the BiTCN model is optimized by replacing the dilated causal convolution with the dilated convolution, an enhanced BiTCN model is constructed, the hyperparameters of the enhanced BiTCN model are initialized, the enhanced BiTCN model replaces the BiTCN model in the BiTCN-GRU model, an improved BiTCN-GRU model is constructed, the multimodal patient rehabilitation feature data is input into the improved BiTCN-GRU model, and the rehabilitation progress prediction result is output; Step S5: Model optimization: In the central processing unit module, an emotion-driven mechanism is introduced to optimize the position update rule of the cat swarm optimization algorithm, an emotion-cat swarm optimization algorithm is constructed, hyperparameters of the improved BiTCN-GRU model are optimized according to the emotion-cat swarm optimization algorithm, an optimal hyperparameter configuration is generated, and the rehabilitation progress prediction result is improved according to the optimal hyperparameter configuration to generate an optimized rehabilitation progress prediction result; Step S6: Visualization report generation.

2. A method for tracking the rehabilitation progress of orthopedic patients according to claim 1, characterized in that: The improved BiTCN-GRU model includes an enhanced BiTCN model, a GRU layer, a Dropout layer, and a fully connected layer to define the rehabilitation progress prediction space.

3. A method for tracking the rehabilitation progress of orthopedic patients according to claim 2, characterized in that: Step S4 specifically includes the following steps: Step S41: extracting local features of multimodal patient rehabilitation feature data through the enhanced BiTCN model to obtain comprehensive time series feature data; Step S42: performing time series analysis on the comprehensive time series feature data through the GRU layer to obtain time-dependent feature data; Step S43: Processing the time-dependent feature data through the Dropout layer, performing regularization processing by randomly discarding some neurons, and obtaining regularized time series feature data; Step S44: Mapping the regularized time series feature data to the rehabilitation progress prediction space through a fully connected layer to obtain a rehabilitation progress prediction result.

4. A method for tracking the rehabilitation progress of orthopedic patients according to claim 3, characterized in that: Step S41 specifically includes the following steps: Step S411: data organization: organizing the multimodal patient rehabilitation feature data into multidimensional time series feature data in chronological order; Step S412: Forward local feature extraction: Process the multidimensional time series feature data through dilated causal convolution, expand the receptive field, extract the local dependency features of the time dimension, obtain dilated convolution feature data, represent the dilated convolution feature data as vectorized capsule data, capture the space-time relationship of the vectorized capsule data through the dynamic routing mechanism of the capsule network, and generate the forward convolution result; Step S413: forward feature processing: batch standardization, weight normalization, nonlinear mapping and regularization are performed on the forward convolution result to extract the historical dependency relationship in the multidimensional time series feature data and generate forward time-dependent feature data; Step S414: reverse local feature extraction: the multidimensional time series feature data is processed by a dilated convolution pyramid, and a dilated convolution with different dilation rates is used to form a pyramid structure to obtain convolution results with different dilation rates, further expand the receptive field, capture the multi-scale time dependency of the multidimensional time series feature data, and fuse the convolution results with different dilation rates to generate a reverse convolution result; Step S414: reverse feature processing: performing batch standardization, weight normalization, nonlinear mapping and regularization processing on the reverse convolution result, extracting future dependency relations in the multidimensional time series feature data, and generating reverse time dependency feature data; Step S414: feature fusion: fusing the forward time-dependent feature data and the reverse time-dependent feature data to generate bidirectional time-dependent feature data; Step S415: Feature adjustment: perform dimension adjustment on the bidirectional time-dependent feature data and convert it into comprehensive time series feature data.

5. A method for tracking the rehabilitation progress of orthopedic patients according to claim 1, characterized in that: Step S5 specifically includes the following steps: Step S51: Initialize the population: randomly generate cat individuals in the population of the cat population optimization algorithm, each cat individual represents a set of hyperparameter configurations of the improved BiTCN-GRU model, define the search space, and generate the initial cat population according to the search space; Step S52: Setting the fitness function: defining the fitness function, calculating the initial fitness value of the initial cat population by the fitness function, and selecting the current global optimal solution and the local optimal solution according to the initial fitness value; Step S53: Position update: introduce an emotion-driven mechanism to simulate the emotional fluctuations of individual cats during the search process, optimize the position update rule of the cat population optimization algorithm, construct an emotion-driven-position update rule, and adjust the position of each individual cat according to the emotion-driven-position update rule. The individual cat updates its position according to the current global optimal solution and the local optimal solution to generate an updated cat population; Step S54: genetic operation: performing crossover operation and mutation operation on the updated cat population to increase the diversity of the population and generate a genetic cat population; calculating the genetic fitness value of the genetic cat population according to the fitness function, and optimizing the current global optimal solution and the local optimal solution according to the genetic fitness value; Step S55: Iterative optimization: set the maximum number of iterations, repeat steps S53 to S54 until the maximum number of iterations is reached, generate a global optimal solution, obtain the optimal hyperparameter configuration, improve the rehabilitation progress prediction result according to the optimal hyperparameter configuration, and generate an optimized rehabilitation progress prediction result.

6. A system for tracking the rehabilitation progress of orthopedic patients, used to implement the above method, characterized in that: The system includes a data acquisition module, a central processing unit module and a visual report generation module; the data acquisition module collects patient rehabilitation data and transmits it to the central processing unit module, the central processing unit module extracts and predicts features of the patient rehabilitation data, and transmits the prediction results to the visual report generation module, and the visual report generation module generates a rehabilitation progress curve, a joint range of motion report and an image change report based on the prediction results.

Images