A method for predicting systemic lupus erythematosus complicated with osteonecrosis based on an improved slime mold optimization algorithm

Abstract

The present application relates to a kind of based on improved slime optimization algorithm's systemic lupus erythematosus concurrent bone necrosis prediction method, belong to medical data processing and disease prediction field.The method includes: collecting SLE patient clinical data and pretreatment;Improved feature selection algorithm of fusion gaussian mixture model and slime algorithm is constructed, the probability distribution modeling of high-quality solution is carried out by gaussian mixture model, and the search process of slime algorithm is integrated, and improved slime optimization algorithm is formed;Continuous value is mapped to binary value using conversion function, to give consideration to the fitness function of classification error rate and feature quantity Key prediction features are selected;Support vector machine is used as classifier, and based on the selected features, a prediction model is constructed to predict bone necrosis risk.The present application can effectively improve the accuracy and efficiency of feature selection, enhance the generalization ability of the model, and provide reliable technical support for early identification and clinical intervention of SLE patients with concurrent bone necrosis.

Description

Technical Field

[0001] This invention relates to the field of medical data processing and disease risk prediction, specifically to a method for predicting osteonecrosis as a complication of systemic lupus erythematosus based on an improved slime mold optimization algorithm. Background Technology

[0002] Systemic lupus erythematosus (SLE) is a chronic autoimmune disease that is prone to osteonecrosis during its course. Osteonecrosis, caused by impaired blood supply to bone tissue leading to bone cell death, has a high disability rate and severely impacts patients' quality of life. Clinical data shows that the prevalence of symptomatic osteonecrosis in SLE patients is approximately 9%, while the prevalence of asymptomatic osteonecrosis in screening is as high as 29%. Therefore, early and accurate identification of high-risk individuals from a wealth of clinical data before irreversible bone collapse occurs is crucial for clinical intervention.

[0003] However, the clinical features of osteonecrosis are nonspecific, and some patients may even be asymptomatic, making it easy to miss the optimal treatment window with traditional diagnosis. Machine learning technology has provided a new approach for predicting osteonecrosis, but clinical data from systemic lupus erythematosus (SLE) suffers from high dimensionality and numerous redundant features: on the one hand, redundant features increase the computational complexity of the model, leading to low training efficiency; on the other hand, irrelevant features can easily cause model overfitting, reducing prediction reliability. Therefore, feature selection has become a core element in improving the predictive performance of osteonecrosis as a complication of SLE.

[0004] Among existing feature selection methods, filtering methods are computationally efficient but ignore the correlation between features, embedded methods are coupled with model training but have high complexity, and wrapping methods are suitable for classification tasks but require traversing feature subsets, resulting in extremely high computational costs. Metaheuristic algorithms are widely used for wrapping feature selection due to their strong global search capabilities, but traditional algorithms suffer from strong search blindness and are prone to getting trapped in local optima, making it difficult to accurately select key features in high-dimensional medical data.

[0005] Therefore, there is an urgent need for an improved algorithm that can balance search efficiency and feature selection accuracy to enhance the accuracy and practicality of predicting osteonecrosis as a complication of systemic lupus erythematosus. Summary of the Invention

[0006] This invention designs a method for predicting osteonecrosis in systemic lupus erythematosus (SLE) based on an improved slime mold optimization algorithm. The technical problem it solves is the poor model performance and difficulty in identifying key features in the existing SLE osteonecrosis prediction method due to feature redundancy.

[0007] To solve the aforementioned technical problems, the present invention adopts the following solution: A method for predicting osteonecrosis as a complication of systemic lupus erythematosus based on an improved slime mold optimization algorithm, characterized by comprising the following steps: Step 1: Data Acquisition and Preprocessing Clinical data of patients with systemic lupus erythematosus were collected, and the clinical data were subjected to missing value imputation, outlier removal and standardization to obtain preprocessed data; the clinical data included clinical symptoms, laboratory test results and medication history, etc. Step 2: Improve the feature selection algorithm construction: By integrating Gaussian mixture model and slime mold algorithm, the probability distribution of high-quality solutions in the algorithm iteration process is modeled by Gaussian mixture model, and the obtained distribution information is incorporated into the search process of slime mold algorithm to obtain an improved slime mold optimization algorithm. The distribution information is incorporated into the search process of the slime mold algorithm, including guiding the slime mold algorithm to focus on feature subsets that conform to the data distribution in the initial search stage, incorporating the likelihood value derived from the Gaussian mixture model into the fitness function in the shrinking and expanding stage, and adjusting the individual movement step size of the slime mold algorithm in combination with the clustering results of the Gaussian mixture model. Step 3, Binary Variant Conversion and Key Feature Selection: A conversion function is used to map the continuous values ​​output by the improved slime mold algorithm to discrete binary values. The constructed fitness function is then used to select features from the preprocessed data to identify key predictive features. Step 4: Predictive Model Construction and Risk Prediction: Using support vector machine as the classifier, a predictive model is constructed based on the selected key features. The preprocessed data of the patients to be predicted is input into the model, and the risk of osteonecrosis is output.

[0008] Preferably, in step 2, the execution process of the improved slime mold optimization algorithm includes: Step 21: Initialize parameters, including population size N, maximum number of iterations MaxIter, and dimension D; Step 22: Initialize population location; Step 23: Execute the dynamic adaptive Gaussian mixture model strategy: Select high-quality solutions in the current population and train the Gaussian mixture model to capture the probability distribution of the optimal solution region; Step 24: Execute the mutation strategy guided by the probability subspace: Construct important subspaces based on the clustering results of the Gaussian mixture model, and update the individual positions according to the current iteration number and the maximum iteration number; Step 25: Execute the Gaussian mixture difference adaptive operator strategy: In the development phase, the idea of ​​difference evolution and the cluster-based adaptive step size mechanism are introduced to update the individual positions; Step 26: Calculate and sort the fitness values ​​of the updated population, and update the global optimal solution using a greedy selection method; Step 27: Repeat steps 23 to 26 until the iteration termination condition is met, and output the optimal position and optimal fitness value.

[0009] Preferably, the dynamic adaptive Gaussian mixture model strategy in step 23 is obtained by maximizing the likelihood function and using the expectation-maximization algorithm to iteratively solve for high-quality solution data. The specific steps are as follows: Step 231, Model Initialization: Randomly initialize the mixing weights α, mean vector μ, and covariance matrix Σ for each Gaussian component; Step 232, Expected Step: Based on the current model parameters, calculate the posterior probability that each high-quality solution belongs to each Gaussian component. ; Step 233, Maximization step: Update the weights, mean vector, and covariance matrix parameters of the Gaussian mixture model using the posterior probability; Step 234, Iteration and Convergence: Repeat the expectation step and the maximization step until the gain of the model's log-likelihood function value is lower than the preset threshold or the maximum number of iterations is reached, to obtain the final Gaussian mixture model parameters.

[0010] Preferably, the posterior probability in step 232 This reflects the probability that a population belongs to each Gaussian component, and its formula is:

[0011] ; in, Indicates that in a given number Given the mean and covariance of each component, a high-quality solution Gaussian probability density function value, The total number of Gaussian distributions pre-defined in the Gaussian mixture model; high-quality solution Let i represent the i-th individual in the set of high-quality solutions used to train the Gaussian mixture model; In step 233, the posterior probabilities calculated in the expectation step of step 232 are used to update all parameters of the model to maximize the log-likelihood function of the complete data. The update formula is as follows:

[0012] ;

[0013] ;

[0014] .

[0015] Preferably, step 22 involves initializing the population location; As shown in the following formula:

[0016] ; in, , It refers to population size; , It refers to the size of the dimension; and They represent the first Maximum and minimum boundaries in each dimension; Used to generate random numbers between 0 and 1; The probability subspace-guided mutation strategy in step 24, based on the clustering results of the Gaussian mixture model, performs eigenvalue decomposition on the covariance matrix of the cluster to which an individual belongs, selects the eigenvectors corresponding to the top t largest eigenvalues ​​to form an important subspace, and then determines the subspace based on the current iteration number. and maximum number of iterations The individual location is updated using the following formula:

[0017] ;

[0018] ;

[0019] ; in, The center of the cluster to which the individual belongs. It is a dynamic shrinkage coefficient used to control the individual update step size. The binary indicator vectors formed by the important subspaces guide the search toward information-rich dimensions. It is a weighting coefficient dynamically calculated based on the ascending order of the fitness values ​​of individuals in the population, so that individuals with better fitness have greater influence during the update process. and These represent the current best fitness and worst fitness, respectively. The binary indicator vector formed by the important subspace is used to binarize the first 5 important feature vectors. If the dimension belongs to the important subspace, the corresponding position is 1, otherwise it is 0.

[0020] Preferably, the hybrid Gaussian difference adaptive operator strategy in step 25 introduces the idea of ​​differential evolution and a cluster-based adaptive step size mechanism during the development stage of the slime mold algorithm, generates variable step size, and adaptively adjusts the shrinkage factor based on the distance from the individual to its cluster center to balance global exploration and local development capabilities. The individual position update formula is as follows:

[0021] ; in, and These are two individuals randomly selected from the current population that are completely different from the current individual. The contraction coefficient is within a dynamically changing range. Take a random value from the inside. The calculation formula is as follows:

[0022] .

[0023] Preferably, step 3 specifically includes: Step 31, Binary Conversion: The individual position vectors output by the improved slime mold algorithm in continuous space are mapped to binary vectors using a conversion function, where "1" indicates selection of the corresponding feature and "0" indicates no selection. The conversion function is as follows:

[0024] ; in, r is a random number in the interval [0,1]. Step 32, Fitness Function Evaluation: Constructing the fitness function ; in, For classification error rate, This represents the individual fitness value. It was set to 0.95. This indicates the number of features selected in the subset. Represents the total number of characteristics; Step 33, Iterative Optimization: Using the fitness function as the optimization objective, the algorithm iteratively searches to find the binary vector that minimizes the fitness function value, which is then used as the optimal feature combination.

[0025] A system for predicting osteonecrosis as a complication of systemic lupus erythematosus based on an improved slime mold optimization algorithm, characterized in that it comprises: The data processing module is used to implement the functions described in step 1 above; The feature selection setting module is used to set the feature selection task and fitness function; The algorithm processing module is used to implement the construction and feature selection functions of the improved slime mold optimization algorithm in steps 2 and 3 above; The model building and prediction module is used to implement the functions in step 4 above.

[0026] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: when the processor executes the computer program, it implements the steps of the above-described method.

[0027] A computer-readable storage medium having a computer program stored thereon, characterized in that, when the computer program is loaded by a processor, it is capable of executing the steps of the method described above.

[0028] The method for predicting osteonecrosis in systemic lupus erythematosus based on an improved slime mold optimization algorithm has the following beneficial effects: (1) This invention constructs an improved slime mold optimization algorithm by integrating Gaussian mixture model and slime mold algorithm. It uses Gaussian mixture model to probabilistically model high-quality solutions, effectively captures the intrinsic structure of data, guides the algorithm to focus on high-potential feature subspace, significantly reduces search blindness, and improves the accuracy and convergence speed of feature selection.

[0029] (2) By adopting a binary conversion mechanism and a comprehensive fitness function, this invention controls the number of features while ensuring prediction accuracy, effectively avoids overfitting, and improves the generalization ability and stability of the prediction model under complex clinical data.

[0030] (3) This invention introduces a probabilistic subspace-guided mutation and a mixture of Gaussian differences adaptive operator strategy, combined with dynamic shrinkage coefficient and weight mechanism, to enhance the adaptive adjustment capability of the algorithm between exploration and development, avoid getting trapped in local optima, and ensure efficient optimization in high-dimensional medical data.

[0031] (4) This invention constructs a prediction model based on support vector machine and combines it with the selected key features to achieve accurate prediction of the risk of osteonecrosis in patients with systemic lupus erythematosus, providing a reliable early screening tool for clinical practice and helping to identify and intervene in high-risk groups.

[0032] (5) The proposed improved slime mold optimization algorithm not only performs well in osteonecrosis prediction, but can also be extended to feature selection and prediction tasks of other high-dimensional medical data, and has good portability and application prospects. Attached Figure Description

[0033] Figure 1 This is a flowchart illustrating a method for predicting osteonecrosis complicated by systemic lupus erythematosus based on an improved slime mold optimization algorithm, according to an embodiment of the present invention.

[0034] Figure 2 This is a graph showing the convergence curve results of the systemic lupus erythematosus complication osteonecrosis prediction method based on the improved slime mold optimization algorithm in this embodiment of the invention, and the results of each algorithm on 20 public datasets.

[0035] Figure 3 This is a graph showing the convergence curve results of the systemic lupus erythematosus (SLE) osteonecrosis prediction method based on the improved slime mold optimization algorithm in this embodiment of the invention, and the results of each algorithm on the SLE osteonecrosis dataset. Detailed Implementation

[0036] The following is combined Figures 1 to 3 The present invention will be further described as follows: like Figure 1As shown, the process of this invention begins with the data acquisition and preprocessing module, followed by the model initialization stage, where algorithm parameters and population are set. The core component is an iterative optimization process based on bGSMA, which nests fitness evaluation and feature selection loops until the termination condition is met, at which point the optimal feature subset is output, and finally, the prediction model is constructed and the risk result is output.

[0037] like Figure 2 As shown, the convergence curves of the bGSMA algorithm proposed in this invention on 20 publicly available medical datasets demonstrate that, compared to traditional optimization algorithms (such as SMA, PSO, etc.), bGSMA has a faster convergence speed and better convergence accuracy. Its curve decreases more rapidly and stabilizes at a lower fitness level, proving the effectiveness of its improved strategy in enhancing search efficiency and avoiding premature convergence.

[0038] like Figure 3 As shown, the bGSMA algorithm demonstrates a more significant performance advantage on a targeted dataset of systemic lupus erythematosus complicated with osteonecrosis. Its convergence curve not only rapidly approaches the global optimum, but its final stable value is also significantly lower than that of the comparative algorithms. This directly verifies that the method of this invention can efficiently screen key predictive features from complex clinical data, providing strong support for the practical clinical application of this method.

[0039] This invention provides a method for predicting osteonecrosis as a complication of systemic lupus erythematosus based on an improved slime mold optimization algorithm, specifically including the following steps: Step 1: Read the medical dataset; Step 2: Divide the dataset into a test set and a training set using 10-fold cross-validation; Step 3: Initialize dimensions within the parameter range Maximum number of iterations Optimal vector set ; Step 4: Initialize the algorithm population; Step 5: Train the training set using a Support Vector Machine (SVM) model and calculate the fitness value for each individual. In each iteration of the bGSMA algorithm, for each individual (position vector) in the population, it is first mapped to a binary feature selection vector using a transformation function. Then, based on the feature subset selected by this vector, a Support Vector Machine (SVM) is trained on the training set, and the classification error rate Err is calculated on an independent validation set (or through cross-validation). The fitness value F of the individual is then calculated in conjunction with the number of features. bGSMA maps the continuous solution space to a discrete feature selection scheme through a transformation function and uses a support vector machine model based on ten-fold cross-validation to evaluate the feature subset. The algorithm calculates the classification error rate on the validation set... Err With the number of features As a joint optimization objective, the optimal feature subset is selected by calculating the fitness value F.

[0040] Calculating the fitness value of an individual using a support vector machine (SVM) model requires converting each individual vector into a binary vector using a transformation function before calculating its fitness value. The specific calculation is as follows:

[0041] ;

[0042] ;

[0043] ; in, To convert function values, Indicates the first Individuals in a population Dimensional value, For classification error rate, This represents the individual fitness value. It was set to 0.95. This indicates the number of features selected in the subset. This represents the total number of characteristics.

[0044] Step 6: Use bGSMA to find the optimal feature subset; Step 7: Determine if the current iteration count exceeds the maximum iteration count. If not, repeat steps 5 and 6.

[0045] Step 8: Output the optimal feature subset.

[0046] Step 6 specifically includes: (a) Model initialization; Mixing weights for each Gaussian component Mean vector Covariance Matrix Perform random initialization; (b) Expected steps; Based on the parameters of the current model, calculate the posterior probability of each high-quality solution generated by each Gaussian component. This probability reflects the likelihood that a population belongs to any of the Gaussian components (clusters), and its formula is:

[0047] ; in, Indicates that in a given number Given the mean and covariance of each component, the population Gaussian probability density function value, This represents the total number of Gaussian distributions preset in the Gaussian mixture model. (c) Maximize the steps; The posterior probabilities calculated in the expectation step are used to update all parameters of the model to maximize the log-likelihood function of the complete data. The update formula is as follows:

[0048] ;

[0049] ;

[0050] ; (d) Iteration and convergence; Repeat the expected step and the maximization step until the gain of the model's log-likelihood function value is lower than a preset threshold or the maximum number of iterations is reached, at which point the model is determined to have converged, thus obtaining the final Gaussian mixture model parameters used to guide the algorithm search.

[0051] (e) Probability subspace-guided mutation strategy; Based on the clustering results of the Gaussian mixture model, eigenvalue decomposition is performed on the covariance matrix of the cluster to which an individual belongs. The eigenvectors corresponding to the top 5 largest eigenvalues ​​are selected to form an important subspace, and then the current iteration number is used as a basis for further analysis. ) and maximum number of iterations ( The individual location is updated using the following formula:

[0052] ;

[0053] ;

[0054] ; in, The center of the cluster to which the individual belongs. It is a dynamic shrinkage coefficient used to control the individual update step size. The binary indicator vector formed by the important subspaces is used to binarize the first 5 selected important feature vectors. If the dimension belongs to an important subspace, the corresponding position is 1; otherwise, it is 0, thus guiding the search towards the information-rich dimensions. It is a weighting coefficient dynamically calculated based on the ascending order of the fitness values ​​of individuals in the population, so that individuals with better fitness have greater influence during the update process. and These represent the current best fitness and worst fitness, respectively. (f) Mixture Gaussian difference adaptive operator strategy; In the development phase of the slime mold algorithm, differential evolution and a cluster-based adaptive step size mechanism are introduced to generate variable step sizes. The shrinkage factor is adaptively adjusted based on the distance from the individual to its cluster center to balance global exploration and local development capabilities. The specific calculation is as follows:

[0055] ; in, and These are two individuals randomly selected from the current population that are completely different from the current individual. The contraction coefficient is within a dynamically changing range. Take a random value from the inside. The calculation formula is as follows:

[0056] ; (g) Transform all individuals using the aforementioned transformation function and calculate their fitness values; This invention performs feature selection on 20 publicly available medical datasets, including publicly available datasets from the UCI Machine Learning Library. Specific data information is shown in Table 1. Each algorithm is run independently 10 times, using 10-fold cross-validation. The dataset is divided into 10 equal parts, one part of which is randomly selected for the test set, and the remaining parts are used for the training set. The number of individuals (N) for all evaluation algorithms is set to 30, and the maximum number of iterations is (…). The value was set to 50. The algorithm was evaluated based on average fitness, average error rate, and average feature count. The Wilcoxon signed-rank test was used to compare the experimental results in terms of fitness, classification error rate, and number of selected features, as shown in Table 2-4. Here, "+ / - / =" indicates that bGSMA is better than the algorithm, worse than the algorithm, and similar to the algorithm, respectively.

[0057] Table 120 datasets details

[0058] Table 2. Fitness ranking of bGSMA on 20 datasets

[0059] Table 3. Error rate ranking of bGSMA on 20 datasets

[0060] Table 4. Feature Count Ranking of bGSMA on 20 Datasets

[0061] In addition to the aforementioned publicly available datasets, to further verify the effectiveness and applicability of the feature selection method proposed in this invention in the diagnosis of complex clinical diseases, this study constructed and used a clinical dataset of systemic lupus erythematosus complicated with osteonecrosis from the National Rheumatology Data Center (CRDC) of Ningbo Medical Center Li Huili Hospital. This dataset included 163 valid samples, all of whom strictly conformed to the classification criteria published by the International Lupus Collaboration Group (SLICC) or ACR / EULAR. Based on clinical manifestations and diagnostic results from X-ray, CT, or MRI imaging, the samples were divided into an osteonecrosis group (52 cases) and a non-osteonecrosis group (111 cases). The ratio of the case group to the control group was approximately 1:2.1, which accurately reflects the distribution characteristics of this complication in actual clinical scenarios while also preserving the class imbalance characteristic common in medical data.

[0062] This invention, based on the aforementioned clinical dataset of systemic lupus erythematosus complicated with osteonecrosis, employs a 10-fold cross-validation scheme to ensure the robustness of the evaluation results. Six multidimensional evaluation metrics—accuracy, sensitivity, specificity, precision, Matthews correlation coefficient (MCC), and F-measure—are selected to comprehensively evaluate the performance of the bGSMA algorithm. The experimental results are shown in Table 5. The experimental statistical results demonstrate that the bGSMA algorithm exhibits excellent overall performance. Specifically, the algorithm achieved an average accuracy of 95.532% and an average sensitivity of 95.555%, ranking first among all compared algorithms, fully demonstrating its powerful ability to correctly identify positive osteonecrosis samples. Simultaneously, the algorithm also ranks first in MCC and F-Measure, reflecting its robustness and balance in handling imbalanced data. Although it ranks fourth in specificity and precision, its values ​​remain at a very high level, demonstrating the algorithm's reliability in reducing the misdiagnosis rate.

[0063] Table 5. Experimental results of the improved slime mold optimization algorithm on a clinical dataset complicated with osteonecrosis.

[0064]

[0065] The outstanding features of this invention compared to existing technologies are: This invention proposes an improved slime mold algorithm, which utilizes multiple strategy mechanisms to enhance the algorithm's exploration capabilities, accelerate convergence speed, strengthen inter-individual correlations, and finally demonstrate the effectiveness of bGSMA in selecting features from medical data of different dimensions in different dimensional medical datasets. It also effectively verifies its predictive performance on a dataset of systemic lupus erythematosus complicated with osteonecrosis.

[0066] The modules in the above-mentioned method for treating systemic lupus erythematosus complicated with osteonecrosis based on the improved slime mold optimization algorithm can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0067] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory. The processor executes the computer program to implement the steps in the embodiment of the method for treating osteonecrosis complicated by lupus erythematosus based on an improved slime mold optimization algorithm. Specific implementation methods can be found in the method embodiments, and will not be repeated here.

[0068] Furthermore, the present invention also provides a non-transitory computer-readable storage medium containing instructions on which a computer program is stored. For example, a memory containing instructions that can be executed by a processor of a computer device to perform the above-described method. For example, the non-transitory computer-readable storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device, etc. When the computer program is executed by the processor, it can implement the steps in the embodiments of the method for predicting osteonecrosis complicated by systemic lupus erythematosus based on an improved slime mold optimization algorithm. Specific implementation methods can be found in the method embodiments, which will not be repeated here.

[0069] Those skilled in the art will understand that embodiments of the present invention can provide methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0070] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0071] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0072] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0073] The present invention has been described above by way of example with reference to the accompanying drawings. Obviously, the implementation of the present invention is not limited to the above-described manner. Any improvements made using the inventive concept and technical solution of the present invention, or the direct application of the inventive concept and technical solution of the present invention to other occasions without modification, are all within the protection scope of the present invention.

Claims

1. A method for predicting osteonecrosis as a complication of systemic lupus erythematosus based on an improved slime mold optimization algorithm, characterized in that, Includes the following steps: Step 1: Data Acquisition and Preprocessing Clinical data of patients with systemic lupus erythematosus were collected, and the clinical data were subjected to missing value imputation, outlier removal and standardization to obtain preprocessed data; Step 2: Improve the feature selection algorithm construction: By integrating Gaussian mixture model and slime mold algorithm, the probability distribution of high-quality solutions in the algorithm iteration process is modeled by Gaussian mixture model, and the obtained distribution information is incorporated into the search process of slime mold algorithm to obtain an improved slime mold optimization algorithm. Step 3, Binary Variant Conversion and Key Feature Selection: A conversion function is used to map the continuous values ​​output by the improved slime mold algorithm to discrete binary values. The constructed fitness function is then used to select features from the preprocessed data to identify key predictive features. Step 4: Predictive Model Construction and Risk Prediction: Using support vector machine as the classifier, a predictive model is constructed based on the selected key features. The preprocessed data of the patients to be predicted is input into the model, and the risk of osteonecrosis is output.

2. The method for predicting osteonecrosis in systemic lupus erythematosus based on an improved slime mold optimization algorithm according to claim 1, characterized in that: In step 2, the execution process of the improved slime mold optimization algorithm includes: Step 21: Initialize parameters, including population size N, maximum number of iterations MaxIter, and dimension D; Step 22: Initialize population location; Step 23: Execute the dynamic adaptive Gaussian mixture model strategy: Select high-quality solutions in the current population and train the Gaussian mixture model to capture the probability distribution of the optimal solution region; Step 24: Execute the mutation strategy guided by the probability subspace: Construct important subspaces based on the clustering results of the Gaussian mixture model, and update the individual positions according to the current iteration number and the maximum iteration number; Step 25: Execute the Gaussian mixture difference adaptive operator strategy: In the development phase, the idea of ​​difference evolution and the cluster-based adaptive step size mechanism are introduced to update the individual positions; Step 26: Calculate and sort the fitness values ​​of the updated population, and update the global optimal solution using a greedy selection method; Step 27: Repeat steps 23 to 26 until the iteration termination condition is met, and output the optimal position and optimal fitness value.

3. The method for predicting osteonecrosis in systemic lupus erythematosus based on an improved slime mold optimization algorithm according to claim 2, characterized in that: The dynamic adaptive Gaussian mixture model strategy in step 23 is obtained by maximizing the likelihood function and using the expectation-maximization algorithm to iteratively solve for high-quality solution data. The specific steps are as follows: Step 231, Model Initialization: Randomly initialize the mixing weights α, mean vector μ, and covariance matrix Σ for each Gaussian component; Step 232, Expected Step: Based on the current model parameters, calculate the posterior probability that each high-quality solution belongs to each Gaussian component. ; Step 233, Maximization step: Update the weights, mean vector, and covariance matrix parameters of the Gaussian mixture model using the posterior probability; Step 234, Iteration and Convergence: Repeat the expectation step and the maximization step until the gain of the model's log-likelihood function value is lower than the preset threshold or the maximum number of iterations is reached, to obtain the final Gaussian mixture model parameters.

4. The method for predicting osteonecrosis in systemic lupus erythematosus based on an improved slime mold optimization algorithm according to claim 3, characterized in that: Posterior probability in step 232 This reflects the probability that a high-quality solution group belongs to each Gaussian component, and its formula is: ； in, Indicates that in the given first Given the mean and covariance of each component, a high-quality solution Gaussian probability density function value, The total number of Gaussian distributions pre-defined in the Gaussian mixture model; high-quality solution Let i represent the i-th individual in the set of high-quality solutions used to train the Gaussian mixture model; In step 233, the posterior probabilities calculated in the expectation step of step 232 are used to update all parameters of the model to maximize the log-likelihood function of the complete data. The update formula is as follows: ； ； 。 5. The method for predicting osteonecrosis in systemic lupus erythematosus based on an improved slime mold optimization algorithm according to claim 2, characterized in that: Step 22: Initialize population location; As shown in the following formula: ； in, , It refers to population size; , It refers to the size of the dimension; and They represent the first Maximum and minimum boundaries in each dimension; Used to generate random numbers between 0 and 1; The probability subspace-guided mutation strategy in step 24, based on the clustering results of the Gaussian mixture model, performs eigenvalue decomposition on the covariance matrix of the cluster to which an individual belongs, selects the eigenvectors corresponding to the top t largest eigenvalues ​​to form an important subspace, and then determines the subspace based on the current iteration number. and maximum number of iterations The individual location is updated using the following formula: ； ； ； Where t is a preset important subspace dimension, and its value is a positive integer. The center of the cluster to which the individual belongs. It is a dynamic shrinkage coefficient used to control the individual update step size. S is a binary indicator vector composed of the important subspaces. The selected top 5 important feature vectors are binarized. If the dimension belongs to the important subspace, the corresponding position is 1; otherwise, it is 0, thereby guiding the search towards the information-rich dimensions. It is a weighting coefficient dynamically calculated based on the ascending order of the fitness values ​​of individuals in the population, so that individuals with better fitness have greater influence during the update process. and These represent the current best fitness and worst fitness, respectively.

6. The method for predicting osteonecrosis in systemic lupus erythematosus based on an improved slime mold optimization algorithm according to claim 2, characterized in that: The hybrid Gaussian difference adaptive operator strategy in step 25 introduces the idea of ​​differential evolution and a cluster-based adaptive step size mechanism during the development stage of the slime mold algorithm. It generates variable step size and adaptively adjusts the shrinkage factor based on the distance from the individual to its cluster center to balance global exploration and local development capabilities. The individual position update formula is as follows: ； in, and These are two individuals randomly selected from the current population that are completely different from the current individual. The contraction coefficient is within a dynamically changing range. Take a random value from the inside. The calculation formula is as follows: 。 7. The method for predicting osteonecrosis in systemic lupus erythematosus based on an improved slime mold optimization algorithm according to claim 2, characterized in that: Step 3 specifically includes: Step 31, Binary Conversion: The individual position vectors output by the improved slime mold algorithm in continuous space are mapped to binary vectors using a conversion function, where "1" indicates that the corresponding feature is selected and "0" indicates that it is not selected. The conversion function is as follows: ； in, r is a random number in the interval [0,1]. Step 32, Fitness Function Evaluation: Constructing the fitness function ; in, For classification error rate, This represents the individual fitness value. It was set to 0.

95. This indicates the number of features selected in the subset. Represents the total number of characteristics; Step 33, Iterative Optimization: Using the fitness function as the optimization objective, the algorithm iteratively searches to find the binary vector that minimizes the fitness function value, which is then used as the optimal feature combination.

8. A system for predicting osteonecrosis as a complication of systemic lupus erythematosus based on an improved slime mold optimization algorithm, characterized in that, include: The data processing module is used to implement the function of step 1 as described in claim 1; The feature selection setting module is used to set the feature selection task and fitness function; The algorithm processing module is used to implement the construction and feature selection functions of the improved slime mold optimization algorithm in steps 2 and 3 of claim 1; The model building and prediction module is used to implement the function of step 4 as described in claim 1.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is loaded by the processor, it is able to perform the steps of the method as described in any one of claims 1 to 7.

Images