Rheumatism patient rehabilitation nursing system based on artificial intelligence

Abstract

The invention relates to the technical field of image data analysis, provides a rheumatism patient rehabilitation nursing system based on artificial intelligence, and aims to optimize the rehabilitation process of a rheumatism patient through accurate image data analysis and personalized nursing plans. The system mainly comprises a data acquisition module, a data preprocessing module, an evaluation module, an optimization hyper-parameter module and a rehabilitation nursing module. An evaluation module of the system carries out deep analysis on image data of a patient based on rotation isovariant convolution and an LSTM model, an accurate rehabilitation evaluation report is generated, and the evaluation accuracy is improved in combination with multi-scale feature extraction and locality optimization; in the optimization hyper-parameter module, hyper-parameters of the system are optimized through an enhanced firefly optimization algorithm, the system performance is further improved, and the accuracy of a rehabilitation evaluation report is ensured; by combining the technology, the system realizes intelligence and precision of rehabilitation nursing of the rheumatism patient, and provides a personalized and scientific nursing scheme for the patient.

Description

Technical Field

[0001] The present invention relates to the technical field of image data analysis, and in particular to a rehabilitation nursing system for rheumatic patients based on artificial intelligence. Background Art

[0002] As the incidence of rheumatic diseases increases year by year, more and more patients require long-term rehabilitation care. Traditional rehabilitation care systems for rheumatic patients usually rely on rule-based image processing algorithms. However, due to the complexity of rheumatic diseases and individual differences between patients, traditional systems are difficult to achieve precision and personalization, and still have the following shortcomings: First, traditional systems are prone to misjudgment or missed diagnosis when dealing with subtle changes in the early stages of lesions, resulting in low accuracy in image analysis. Subtle changes in image data such as narrowing of joint spaces and bone erosion are difficult to be effectively captured in traditional systems. Second, traditional systems lack a dynamic adjustment mechanism. The system reacts slowly to changes in the patient's condition, making it difficult to adjust the care plan in the first place, resulting in limited efficiency in rehabilitation effects. Therefore, there is an urgent need for a technical solution that can overcome the shortcomings of existing rehabilitation care systems for rheumatic patients and achieve more precise and personalized care. Summary of the invention

[0003] The present invention provides a rehabilitation and nursing system for rheumatic patients based on artificial intelligence, aiming to solve the shortcomings of traditional rehabilitation and nursing systems for rheumatic patients in the prior art and improve the accuracy and effect of rehabilitation and nursing for rheumatic patients; the system adopts a deep learning model based on convolutional neural networks and long short-term memory networks, and combines rotational equivariant convolution, multi-scale feature extraction and convolution local optimization technology to efficiently extract and analyze the medical imaging data of rheumatic patients; through comprehensive analysis of image features and clinical data, the system can comprehensively evaluate the patient's condition and rehabilitation progress, and generate a detailed rehabilitation evaluation report; in order to further improve the prediction accuracy of the system, the present invention introduces an enhanced version of the firefly optimization algorithm, which uses the strategies of nonlinear attraction calculation, dynamic attraction attenuation and mutation operation to finely adjust the parameters of the model, thereby optimizing the performance of the system; through the above-mentioned technical means, the present invention can provide more accurate, personalized and efficient rehabilitation and nursing support for rheumatic patients.

[0004] The present invention provides a rehabilitation nursing system for rheumatoid arthritis patients based on artificial intelligence, the system comprising a data acquisition module, a data preprocessing module, an evaluation module, a hyperparameter optimization module and a rehabilitation nursing module;

[0005] The data acquisition module collects MRI, X-ray, and CT imaging data of rheumatic patients' joints as original rheumatic patient imaging data; collects clinical data for comprehensive analysis, including blood tests and patient symptoms;

[0006] The data preprocessing module performs data enhancement and format conversion on the original rheumatism patient image data to generate preprocessed rheumatism patient image data; data enhancement includes random rotation to avoid CNN being insensitive to rotation; random cropping and flipping to improve generalization ability; Gaussian filtering and contrast enhancement to enhance edge features;

[0007] The evaluation module combines rotational equivariant convolution, multi-scale feature extraction, convolution local optimization and LSTM to construct a symmetric rotational equivariant convolution-LSTM model; initializes the hyperparameters of the symmetric rotational equivariant convolution-LSTM model; analyzes and pre-processes rheumatic patient imaging data through the symmetric rotational equivariant convolution-LSTM model to generate a patient rehabilitation evaluation report;

[0008] The hyperparameter optimization module introduces the firefly optimization algorithm, optimizes the position update strategy of the firefly optimization algorithm through nonlinear attraction calculation, dynamic attraction attenuation, mutation operation, local search strategy and random disturbance control, builds an enhanced version of the firefly optimization algorithm, and uses the enhanced version of the firefly optimization algorithm to optimize the hyperparameters of the symmetric rotation equivariant convolution-LSTM model to obtain the optimal hyperparameter combination; based on the optimal hyperparameter combination, generates an optimized patient rehabilitation assessment report;

[0009] Rehabilitation nursing module, based on the optimized patient rehabilitation assessment report, develops a personalized rehabilitation training plan to optimize the rehabilitation effect of rheumatic patients;

[0010] The symmetric rotation equivariant convolution-LSTM model includes an initialization unit, a binary index matrix generation unit, a rotation equivariant convolution kernel calculation unit, a convolution operation unit, a multi-layer convolution superposition unit, a time series analysis unit, a rehabilitation progress prediction unit and a fully connected layer;

[0011] The enhanced firefly optimization algorithm includes the following steps:

[0012] Step S1: Initialize and calculate brightness;

[0013] Step S2: Calculate the distance and update the position;

[0014] Step S3: Iterate and optimize and select the optimal solution.

[0015] Furthermore, the evaluation module generates a patient rehabilitation evaluation report, which specifically includes the following steps:

[0016] Step D1: The initialization unit sets the size of the convolution kernel and the number of input and output channels, which is used to control the parameter scale of the convolution layer and the complexity of the network structure; non-overlapping symmetric rings are divided according to the size of the convolution kernel to obtain the symmetric structure of the convolution kernel, and the number of symmetric rings is calculated;

[0017] Step D2: The binary index matrix generation unit calculates the Euclidean distance matrix of the convolution kernel according to the size of the convolution kernel and the number of symmetric rings, and uses the locality of the convolution to generate a binary index matrix based on the Euclidean distance matrix to constrain the symmetric structure of the convolution kernel. The formula used is as follows:

[0018] ;

[0019] in, represents pixel coordinates, Represents the convolution kernel position coordinates; Indicates location The local Euclidean distance of Represents the position in the convolution kernel The weight value of Represents pixel coordinates and the convolution kernel position The Euclidean distance between

[0020] Step D3: The rotation equivariant convolution kernel calculation unit defines a parameter matrix according to the symmetric structure of the convolution kernel and the number of symmetric rings. The parameter matrix contains the feature weights of each ring. The rotation equivariant convolution kernel is calculated through the parameter matrix and the binary index matrix.

[0021] Step D4: The convolution operation unit performs a convolution operation on the pre-processed rheumatoid arthritis patient image data using a rotation equivariant convolution kernel, focusing on key area features, including joint space features and bone erosion features; and generates preliminary rotation equivariant feature data;

[0022] Step D5: The multi-layer convolution superposition unit specifically includes: extracting feature data of different scales of preliminary rotation equivariant feature data through three rotation equivariant convolution layers to generate SRE convolution feature maps; then performing pooling processing on the SRE convolution feature maps to reduce the spatial dimension and generate multi-scale SRE convolution feature maps; three rotation equivariant convolution layers: the first convolution kernel is 5×5, the second convolution kernel is 7×7, and the third convolution kernel is 9×9;

[0023] Step D6: The time series analysis unit establishes an LSTM model, inputs clinical data and multi-scale SRE convolutional feature maps into the LSTM model for time series analysis, observes the recovery trend, and generates time series feature data;

[0024] Step D7: The rehabilitation progress prediction unit inputs the time series feature data into the fully connected layer to perform rehabilitation progress prediction and generate a patient rehabilitation assessment report; the patient rehabilitation assessment report includes disease stage classification results and rehabilitation trend analysis data.

[0025] Furthermore, step S1 is specifically as follows: setting the objective function, initializing all firefly individuals in the enhanced firefly optimization algorithm, the position of each firefly individual represents a hyperparameter combination of a symmetric rotation equivariant convolution-LSTM model, calculating the brightness of each firefly individual according to the objective function, and generating a firefly brightness data set.

[0026] Furthermore, step S2 specifically includes:

[0027] Step S21: Calculating the firefly distance between individual fireflies by using the Euclidean distance;

[0028] Step S22: introducing nonlinear dependence and dynamic adjustment mechanism of brightness, and calculating firefly attraction in combination with firefly distance and brightness of individual fireflies;

[0029] Step S23: introducing mutation operation and local search strategy, combining firefly attraction to control the movement of individual fireflies, and updating the firefly brightness dataset;

[0030] The formula used is as follows:

[0031] Attraction attenuation coefficient:

[0032] ;

[0033] in, Represents the iteration index, represents the initial attraction attenuation coefficient, represents the factor that controls the decay rate; Indicated in The attraction decay coefficient in iterations;

[0034] Nonlinear attraction calculation formula:

[0035] ;

[0036] in, represents the individual index of fireflies, Firefly Fireflies The attraction, represents the initial attraction constant, Firefly and fireflies The square of the Euclidean distance between Firefly and fireflies The attraction attenuation factor between represents the weight factor that controls the effect of brightness on attractiveness, represents the nonlinear index controlling the effect of brightness on attractiveness; Firefly Brightness, Firefly Brightness, Firefly and fireflies The brightness ratio of

[0037] The firefly position update formula:

[0038] ;

[0039] in, Firefly Updated location, Firefly Current location; represents the random step coefficient, represents normally distributed random noise, used to introduce randomness; represents the coefficient of variation, represents the random perturbation term used for mutation, represents the local search step coefficient, Represents the objective function The gradient of represents a mutation operation, Represents gradient optimization.

[0040] Furthermore, step S3 is specifically as follows: setting a maximum number of iterations, iteratively updating the firefly brightness data set until the maximum number of iterations is reached, selecting the firefly individual with the best brightness in the firefly brightness data set, and obtaining the optimal hyperparameter combination.

[0041] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0042] The present invention combines rotational equivariant convolution, multi-scale feature extraction, convolution local optimization and LSTM to construct a symmetric rotational equivariant convolution-LSTM model, thereby achieving accurate evaluation of rheumatic patients' imaging data; the evaluation module of the system can comprehensively analyze the pre-processed imaging data, and generate a detailed rehabilitation evaluation report in combination with clinical data; by introducing rotational equivariant convolution and multi-scale feature extraction, the model effectively captures the key features of rheumatic patients' joint images, especially the details of joint space and bone erosion, thereby improving the accuracy of rehabilitation evaluation; the introduction of LSTM makes the prediction of rehabilitation trends more accurate, thereby providing a reliable basis for the formulation of personalized rehabilitation plans; through this technical means, the rehabilitation effect of patients can be more accurately evaluated and predicted;

[0043] In addition, the hyperparameter optimization module of the present invention optimizes the hyperparameters of the symmetric rotation equivariant convolution-LSTM model through an enhanced version of the firefly optimization algorithm, further improving the performance of the evaluation model in the system; by using optimization strategies such as nonlinear attraction calculation, dynamic attraction attenuation, mutation operation, local search strategy and random disturbance control, it is possible to find the optimal hyperparameter combination in a wider search space, so that the performance of the symmetric rotation equivariant convolution-LSTM model is optimal; this method optimizes the learning process of the system, making it converge more efficiently, and avoiding the common local optimal solution problem in traditional systems; through the continuous optimization of hyperparameters, the accuracy of the rehabilitation evaluation report is significantly improved, thereby providing strong support for the formulation of personalized rehabilitation plans for rheumatic patients;

[0044] The present invention combines a variety of cutting-edge technologies, including rotational equivariant convolution, multi-scale feature extraction, LSTM and optimized hyperparameter modules, to achieve highly intelligent and precise image data analysis and rehabilitation assessment processes for rheumatic patients; the application of these technologies not only enhances the personalization and accuracy of rehabilitation care, but also effectively improves the rehabilitation effect of patients; through precise rehabilitation assessment, the recovery process of rheumatic patients has been optimized, and the adjustment of personalized care plans has become more scientific and flexible. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 This is a module schematic diagram of an artificial intelligence-based rehabilitation nursing system for rheumatic patients proposed by the present invention;

[0046] Figure 2 This is a schematic diagram of the initialization process of the rotational equivariant convolution kernel in Example 2. DETAILED DESCRIPTION

[0047] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0048] Embodiment 1, according to Figure 1 , the present invention provides a rehabilitation nursing system for rheumatoid patients based on artificial intelligence, the system comprising a data acquisition module, a data preprocessing module, an evaluation module, a hyperparameter optimization module and a rehabilitation nursing module;

[0049] The data acquisition module collects MRI, X-ray, and CT imaging data of rheumatic patients' joints as original rheumatic patient imaging data; collects clinical data for comprehensive analysis, including blood tests and patient symptoms;

[0050] The data preprocessing module performs data enhancement and format conversion on the original rheumatism patient image data to generate preprocessed rheumatism patient image data; data enhancement includes random rotation to avoid CNN being insensitive to rotation; random cropping and flipping to improve generalization ability; Gaussian filtering and contrast enhancement to enhance edge features;

[0051] The evaluation module combines rotational equivariant convolution, multi-scale feature extraction, convolution local optimization and LSTM to construct a symmetric rotational equivariant convolution-LSTM model; initializes the hyperparameters of the symmetric rotational equivariant convolution-LSTM model; analyzes and pre-processes rheumatic patient imaging data through the symmetric rotational equivariant convolution-LSTM model to generate a patient rehabilitation evaluation report;

[0052] The hyperparameter optimization module introduces the firefly optimization algorithm, optimizes the position update strategy of the firefly optimization algorithm through nonlinear attraction calculation, dynamic attraction attenuation, mutation operation, local search strategy and random disturbance control, builds an enhanced version of the firefly optimization algorithm, and uses the enhanced version of the firefly optimization algorithm to optimize the hyperparameters of the symmetric rotation equivariant convolution-LSTM model to obtain the optimal hyperparameter combination; based on the optimal hyperparameter combination, generates an optimized patient rehabilitation assessment report;

[0053] Rehabilitation nursing module, based on the optimized patient rehabilitation assessment report, develops a personalized rehabilitation training plan to optimize the rehabilitation effect of rheumatic patients;

[0054] The symmetric rotation equivariant convolution-LSTM model includes an initialization unit, a binary index matrix generation unit, a rotation equivariant convolution kernel calculation unit, a convolution operation unit, a multi-layer convolution superposition unit, a time series analysis unit, a rehabilitation progress prediction unit and a fully connected layer;

[0055] The enhanced firefly optimization algorithm includes the following steps:

[0056] Step S1: Initialize and calculate brightness;

[0057] Step S2: Calculate the distance and update the position;

[0058] Step S3: Iterate and optimize and select the optimal solution.

[0059] Embodiment 2, according to Figure 2 This embodiment is based on the first embodiment. In this embodiment, the evaluation module generates a patient rehabilitation evaluation report, which specifically includes the following steps:

[0060] Step D1: The initialization unit sets the size of the convolution kernel and the number of input and output channels, which is used to control the parameter scale of the convolution layer and the complexity of the network structure; non-overlapping symmetric rings are divided according to the size of the convolution kernel to obtain the symmetric structure of the convolution kernel, and the number of symmetric rings is calculated;

[0061] Step D2: The binary index matrix generation unit calculates the Euclidean distance matrix of the convolution kernel according to the size of the convolution kernel and the number of symmetric rings, and uses the locality of the convolution to generate a binary index matrix based on the Euclidean distance matrix to constrain the symmetric structure of the convolution kernel. The formula used is as follows:

[0062] ;

[0063] in, represents pixel coordinates, Represents the convolution kernel position coordinates; Indicates location The local Euclidean distance of Represents the position in the convolution kernel The weight value of Represents pixel coordinates and the convolution kernel position The Euclidean distance between

[0064] Step D3: The rotation equivariant convolution kernel calculation unit defines a parameter matrix according to the symmetric structure of the convolution kernel and the number of symmetric rings. The parameter matrix contains the feature weights of each ring. The rotation equivariant convolution kernel is calculated through the parameter matrix and the binary index matrix.

[0065] Step D4: The convolution operation unit performs a convolution operation on the pre-processed rheumatoid arthritis patient image data using a rotation equivariant convolution kernel, focusing on key area features, including joint space features and bone erosion features; and generates preliminary rotation equivariant feature data;

[0066] Step D5: The multi-layer convolution superposition unit extracts feature data of different scales of the preliminary rotation equivariant feature data through three rotation equivariant convolution layers to generate an SRE convolution feature map; then the SRE convolution feature map is pooled to reduce the spatial dimension and generate a multi-scale SRE convolution feature map; three rotation equivariant convolution layers: the first convolution kernel is 5×5, the second convolution kernel is 7×7, and the third convolution kernel is 9×9;

[0067] Step D6: The time series analysis unit establishes an LSTM model, inputs clinical data and multi-scale SRE convolutional feature maps into the LSTM model for time series analysis, observes the recovery trend, and generates time series feature data;

[0068] Step D7: The rehabilitation progress prediction unit inputs the time series feature data into the fully connected layer to perform rehabilitation progress prediction and generate a patient rehabilitation assessment report; the patient rehabilitation assessment report includes disease stage classification results and rehabilitation trend analysis data.

[0069] Embodiment 3: This embodiment is based on embodiment 1. In this embodiment, the evaluation module generates a patient rehabilitation evaluation report, which specifically includes the following steps:

[0070] Step T1: Set the size of the convolution kernel and the number of input and output channels to control the parameter scale of the convolution layer and the complexity of the network structure; divide the non-overlapping symmetric rings according to the size of the convolution kernel to obtain the symmetric structure of the convolution kernel and calculate the number of symmetric rings;

[0071] Step T2: Calculate the Euclidean distance matrix of the convolution kernel according to the size of the convolution kernel and the number of symmetric rings, generate a binary index matrix according to the Euclidean distance matrix, and constrain the symmetric structure of the convolution kernel;

[0072] Step T3: Define the parameter matrix according to the symmetric structure of the convolution kernel and the number of symmetric rings. The parameter matrix contains the feature weights of each ring. The rotation equivariant convolution kernel is calculated through the parameter matrix and the binary index matrix.

[0073] Step T4: using a rotation equivariant convolution kernel to perform a convolution operation on the preprocessed rheumatic patient image data, focusing on key area features, including joint space features and bone erosion features; generating rotation equivariant feature data;

[0074] Step T5: Establish an LSTM model, input clinical data and transformation feature data into the LSTM model for time series analysis, observe the recovery trend, and generate time series feature data;

[0075] Step T6: Input the time series feature data into the fully connected layer to predict the rehabilitation progress and generate a patient rehabilitation assessment report; the patient rehabilitation assessment report includes the disease stage classification results and rehabilitation trend analysis data.

[0076] Embodiment 4: This embodiment is based on embodiment 1. In this embodiment, the process of generating a patient rehabilitation assessment report by the assessment module specifically includes the following steps:

[0077] Step Q1: performing a convolution operation on the pre-processed rheumatic patient image data to generate convolution feature data;

[0078] Step Q2: Establish an LSTM model, input clinical data and convolution feature data into the LSTM model for time series analysis, observe the recovery trend, and generate time series feature data;

[0079] Step Q3: Input the time series feature data into the fully connected layer to predict the rehabilitation progress and generate a patient rehabilitation assessment report; the patient rehabilitation assessment report includes the disease stage classification results and rehabilitation trend analysis data.

[0080] Embodiment 5. This embodiment is based on embodiment 2. In this embodiment, step S1 is specifically as follows: setting the objective function, initializing all firefly individuals in the enhanced firefly optimization algorithm, the position of each firefly individual represents a hyperparameter combination of a symmetric rotation equivariant convolution-LSTM model, calculating the brightness of each firefly individual according to the objective function, and generating a firefly brightness data set.

[0081] Embodiment 6: This embodiment is based on embodiment 5. In this embodiment, step S2 specifically includes:

[0082] Step S21: Calculating the firefly distance between individual fireflies by using the Euclidean distance;

[0083] Step S22: introducing nonlinear dependence and dynamic adjustment mechanism of brightness, and calculating firefly attraction by combining firefly distance and brightness of individual fireflies;

[0084] Step S23: introducing mutation operation and local search strategy, combining firefly attraction to control the movement of individual fireflies, and updating the firefly brightness dataset;

[0085] The formula used is as follows:

[0086] Attraction attenuation coefficient:

[0087] ;

[0088] in, Represents the iteration index, represents the initial attraction attenuation coefficient, represents the factor that controls the decay rate; Indicated in The attraction decay coefficient in the iteration;

[0089] Nonlinear attraction calculation formula:

[0090] ;

[0091] in, Represents the individual index of fireflies, Firefly Fireflies The attraction, represents the initial attraction constant, Firefly and fireflies The square of the Euclidean distance between Firefly and fireflies The attraction attenuation factor between represents the weight factor that controls the effect of brightness on attractiveness, represents the nonlinear index controlling the effect of brightness on attractiveness; Firefly Brightness, Firefly Brightness, Firefly and fireflies The brightness ratio of

[0092] The firefly position update formula:

[0093] ;

[0094] in, Firefly Updated location, Firefly The current position of , that is, the current hyperparameter combination; represents the random step coefficient, represents normally distributed random noise, used to introduce randomness; represents the coefficient of variation, represents the random perturbation term used for mutation, represents the local search step coefficient, Represents the objective function The gradient of represents a mutation operation, Represents gradient optimization.

[0095] Embodiment 7. This embodiment is based on embodiment 5. In this embodiment, step S2 is specifically as follows: calculating the firefly distance between individual fireflies by Euclidean distance; calculating the firefly attraction by combining the firefly distance and the brightness of the individual fireflies; controlling the movement of the individual fireflies by the firefly attraction, and updating the firefly brightness data set.

[0096] Embodiment 8. This embodiment is based on embodiment 7. In this embodiment, step S3 is specifically as follows: setting a maximum number of iterations, iteratively updating the firefly brightness data set until the maximum number of iterations is reached, selecting the firefly individual with the best brightness in the firefly brightness data set, and obtaining the optimal hyperparameter combination.

[0097] The present invention and its embodiments are described above, which is not restrictive. What is shown in the accompanying drawings is only one of the embodiments of the present invention, and the actual structure is not limited to this. In short, if ordinary technicians in this field are inspired by it and do not deviate from the purpose of the invention, they can creatively design structural methods and embodiments similar to the technical solution, which should all fall within the scope of protection of the present invention.

Claims

1. A rheumatic patient rehabilitation nursing system based on artificial intelligence, comprising a data acquisition module and a data preprocessing module; the data acquisition module acquires original rheumatic patient image data and clinical data; the data preprocessing module preprocesses the original rheumatic patient image data to generate preprocessed rheumatic patient image data; characterized in that: The system also includes an evaluation module and an optimization hyperparameter module; The evaluation module combines rotational equivariant convolution, multi-scale feature extraction, convolution locality optimization and LSTM to construct a symmetric rotational equivariant convolution-LSTM model; initializes the hyperparameters of the symmetric rotational equivariant convolution-LSTM model; analyzes and pre-processes rheumatic patient imaging data through the symmetric rotational equivariant convolution-LSTM model to generate a patient rehabilitation evaluation report; The hyperparameter optimization module introduces the firefly optimization algorithm, optimizes the position update strategy of the firefly optimization algorithm through nonlinear attraction calculation, dynamic attraction attenuation, mutation operation, local search strategy and random disturbance control, constructs an enhanced version of the firefly optimization algorithm, and uses the enhanced version of the firefly optimization algorithm to optimize the hyperparameters of the symmetric rotation equivariant convolution-LSTM model to obtain the optimal hyperparameter combination; based on the optimal hyperparameter combination, an optimized patient rehabilitation assessment report is generated.

2. According to claim 1, a rheumatic patient rehabilitation nursing system based on artificial intelligence is characterized by: The symmetric rotation equivariant convolution-LSTM model includes an initialization unit, a binary index matrix generation unit, a rotation equivariant convolution kernel calculation unit, a convolution operation unit, a multi-layer convolution superposition unit, a time series analysis unit, a rehabilitation progress prediction unit and a fully connected layer.

3. The artificial intelligence-based rehabilitation nursing system for rheumatism patients according to claim 1 is characterized by: The enhanced firefly optimization algorithm includes the following steps: Step S1: Initialize and calculate brightness; Step S2: Calculate the distance and update the position; Step S3: Iterate and optimize and select the optimal solution.

4. The artificial intelligence-based rehabilitation nursing system for rheumatism patients according to claim 2 is characterized by: The process of generating a patient rehabilitation assessment report by the assessment module specifically includes the following steps: Step D1: The initialization unit sets the size of the convolution kernel, divides the convolution kernel into non-overlapping symmetric rings according to the size of the convolution kernel, obtains the symmetric structure of the convolution kernel, and calculates the number of symmetric rings; Step D2: The binary index matrix generation unit calculates the Euclidean distance matrix of the convolution kernel according to the size of the convolution kernel and the number of symmetric rings and utilizes the locality of the convolution, generates a binary index matrix according to the Euclidean distance matrix, and constrains the symmetric structure of the convolution kernel; Step D3: The rotation equivariant convolution kernel calculation unit defines a parameter matrix according to the symmetric structure of the convolution kernel and the number of symmetric rings, and calculates the rotation equivariant convolution kernel through the parameter matrix and the binary index matrix; Step D4: the convolution operation unit performs a convolution operation on the pre-processed rheumatism patient image data using a rotation equivariant convolution kernel to generate preliminary rotation equivariant feature data; Step D5: The multi-layer convolution superposition unit extracts feature data of different scales of the preliminary rotation equivariant feature data through three rotation equivariant convolution layers to generate an SRE convolution feature map; then the SRE convolution feature map is pooled to reduce the spatial dimension and generate a multi-scale SRE convolution feature map; Step D6: The time series analysis unit establishes an LSTM model, inputs the clinical data and the multi-scale SRE convolutional feature map into the LSTM model for time series analysis, and generates time series feature data; Step D7: The rehabilitation progress prediction unit inputs the time series feature data into the fully connected layer to perform rehabilitation progress prediction and generate a patient rehabilitation assessment report.

5. According to claim 3, the artificial intelligence-based rehabilitation nursing system for rheumatism patients is characterized by: Step S1 is specifically as follows: setting the objective function, initializing all firefly individuals in the enhanced firefly optimization algorithm, the position of each firefly individual represents a hyperparameter combination of a symmetric rotation equivariant convolution-LSTM model, calculating the brightness of each firefly individual according to the objective function, and generating a firefly brightness dataset.

6. The artificial intelligence-based rehabilitation nursing system for rheumatoid arthritis patients according to claim 5, characterized in that: Step S2 specifically includes: Step S21: Calculating the firefly distance between individual fireflies by using the Euclidean distance; Step S22: introducing nonlinear dependence and dynamic adjustment mechanism of brightness, and calculating firefly attraction in combination with firefly distance and brightness of individual fireflies; Step S23: Introduce mutation operation and local search strategy, combine firefly attraction to control the movement of individual fireflies, and update the firefly brightness dataset.

7. The artificial intelligence-based rehabilitation nursing system for rheumatism patients according to claim 6 is characterized by: Step S3 is specifically as follows: setting a maximum number of iterations, iteratively updating the firefly brightness data set until the maximum number of iterations is reached, selecting the firefly individual with the best brightness in the firefly brightness data set, and obtaining the optimal hyperparameter combination.

Images