Method and device for predicting occurrence risk of sarcopenia of old people and medium

Abstract

The invention relates to an old people sarcopenia occurrence risk prediction method and device and a medium. The old people sarcopenia occurrence risk prediction method comprises the steps of obtaining sampling indexes of an object to be subjected to old people sarcopenia occurrence risk prediction in multiple dimensions; and inputting the sampling index corresponding to each dimension into a corresponding sub-model in a sarcopenia occurrence risk prediction model to obtain a predicted sarcopenia occurrence result corresponding to a risk prediction object output by the sarcopenia occurrence risk prediction model, the sarcopenia occurrence risk prediction model performs target feature screening on a plurality of sub-models for corresponding sampling indexes, and outputs target features to a multi-factor prediction model result for comprehensive analysis to obtain a predicted sarcopenia occurrence result; wherein the sarcopenia occurrence risk prediction model is obtained by performing sub-model training of corresponding dimensions through sample sampling indexes of multiple dimensions, performing analysis through a multi-element prediction model, performing multiple prediction performance evaluation, and performing verification through a verification set in a sarcopenia queue database.

Description

Technical Field

[0001] The present disclosure relates to the field of medical technology, and in particular to a method, device and medium for predicting the risk of sarcopenia in the elderly. Background Art

[0002] Sarcopenia, as an age-related systemic skeletal muscle disease, has received increasing attention due to its serious impact on the quality of life, independence and survival prognosis of the elderly. Sarcopenia not only leads to decreased physical activity and deterioration of quality of life, but also significantly increases the risk of falls, hospitalization and death, posing a huge economic and care pressure on families and society.

[0003] Therefore, there is an urgent need to provide a method to predict the risk of sarcopenia in the elderly so as to more comprehensively and accurately identify elderly patients at high risk of sarcopenia. Summary of the invention

[0004] The purpose of the present invention is to provide a method, device and medium for predicting the risk of sarcopenia in the elderly, aiming to accurately predict the risk of sarcopenia in the elderly, thereby achieving efficient early detection, early screening and early intervention of sarcopenia in the elderly.

[0005] In order to achieve the above-mentioned purpose, a first aspect of the embodiments of the present disclosure provides a method for predicting the risk of sarcopenia in the elderly, the method comprising: Acquire sampling indicators in multiple dimensions for the subject to be predicted for the risk of sarcopenia in the elderly, wherein the dimensions of the sampling indicators are obtained by summarizing the predicted sampling indicators; Inputting the sampling index corresponding to each dimension into the corresponding sub-model in the sarcopenia occurrence risk prediction model to obtain the predicted sarcopenia occurrence result corresponding to the risk prediction object output by the sarcopenia occurrence risk prediction model, wherein the sarcopenia occurrence risk prediction model screens the target features of multiple sub-models according to the corresponding sampling indexes, and then inputs the target features into the multi-factor prediction model results for comprehensive analysis to obtain the predicted sarcopenia occurrence result; Among them, the sarcopenia risk prediction model is obtained by training the sub-models of corresponding dimensions through sample sampling indicators of multiple dimensions, analyzing through a multi-element prediction model and through multiple prediction performance evaluations, and then verifying through a validation set in a sarcopenia cohort database. The variance inflation factor is used to evaluate the multicollinearity between the sample sampling indicators to ensure the stability of the sarcopenia risk prediction model. The Bootstrap repeated sampling method is used to simulate various sample distribution scenarios, and the performance parameters of the sarcopenia risk prediction model are repeatedly calculated, and consistent calibration and discrimination indicators are used for performance evaluation.

[0006] According to a second aspect of the present disclosure, an electronic device is provided, including: a memory having a computer program stored thereon; A processor is used to execute the computer program in the memory to implement the steps of any one of the methods in the first aspect.

[0007] According to a third aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided, on which a computer program is stored, and when the program is executed by a processor, the steps of any one of the methods described in the first aspect are implemented.

[0008] The present invention provides a method, device and medium for predicting the risk of sarcopenia in the elderly. Compared with the prior art, it has the following beneficial effects: By obtaining the sampling indicators of the elderly in multiple dimensions, the health status of the elderly can be more comprehensively reflected, thereby improving the accuracy of the risk prediction of sarcopenia. The sampling indicators corresponding to each dimension are input into the corresponding sub-model, and the sub-model performs target feature screening, which helps to identify the key features closely related to the occurrence of sarcopenia and reduce the interference of noise data. The target features of multiple sub-models are output to the multi-factor prediction model for comprehensive analysis, which can comprehensively consider the influencing factors in different dimensions and further improve the accuracy and reliability of the prediction. The variance inflation factor is used to evaluate the multicollinearity between the sample sampling indicators, which helps to ensure the stability of the model; the Bootstrap repeated sampling method is used to simulate a variety of sample distribution scenarios, repeatedly calculate the model performance parameters, and use consistent calibration and discrimination indicators for performance evaluation, which can comprehensively evaluate the prediction performance of the model and ensure the accuracy and reliability of the model. The validation set in the sarcopenia cohort database can further verify the prediction effect of the model and ensure the feasibility of the model in practical application. In summary, by obtaining the sampling indicators of the elderly in multiple dimensions, combining the target feature screening of multiple sub-models and the comprehensive analysis of the multi-factor prediction model, the risk of sarcopenia in the elderly can be accurately predicted. At the same time, through rigorous model stability and performance evaluation and validation set verification, the accuracy and reliability of the model are ensured, thereby accurately predicting the risk of sarcopenia in the elderly, thereby achieving efficient early detection, early screening and early intervention of sarcopenia in the elderly.

[0009] Other features and advantages of the present disclosure will be described in detail in the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The accompanying drawings are used to provide a further understanding of the present disclosure and constitute a part of the specification. Together with the following specific embodiments, they are used to explain the present disclosure but do not constitute a limitation of the present disclosure. In the accompanying drawings: Figure 1It is a flow chart of a method for predicting the risk of sarcopenia in the elderly according to an embodiment of the specification.

[0011] Figure 2 It is a flowchart of another method for predicting the risk of sarcopenia in the elderly according to an embodiment of the specification.

[0012] Figure 3 It is a schematic diagram showing an example of evaluating the distinguishing ability of a model by the area under the ROC curve (AUC) according to an embodiment of the specification.

[0013] Figure 4 It is a block diagram of a device for predicting the risk of sarcopenia in the elderly according to an embodiment of the specification. DETAILED DESCRIPTION

[0014] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0015] The specific implementation of the present disclosure is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described herein is only used to illustrate and explain the present disclosure, and is not used to limit the present disclosure.

[0016] Before introducing the prediction method for the risk of sarcopenia in the elderly provided by the present disclosure, we first introduce the technology in the relevant scenarios. For the existing prediction models for sarcopenia in the elderly, after comprehensive systematic evaluation and quality assessment, 18 original literature related to the sarcopenia prediction model were identified. The PROBAST tool was further used to evaluate the bias risk and applicability of the 18 original studies included. In the evaluation of the overall bias risk, the vast majority of studies (n=17, 94.4%) were rated as high bias risk, and only 1 study (accounting for 5.6%) was rated as low bias risk. In terms of the overall applicability assessment, a small number of studies (22.2%, a total of 4) were considered to have a high risk of bias, while most studies (77.8%, a total of 14) were rated as low bias risk.

[0017] Overall, there were large differences in the results of different studies, and according to the assessment of PROBAST bias risk tool, these studies had a high risk of bias. The included models had limitations in methodological quality, which may limit their performance and the breadth of practical application.

[0018] Furthermore, in the relevant scenarios, multivariate logistic analysis was used to explore the risk factors for sarcopenia in 1,050 elderly people in the community of Hunan, China. The results showed that 6 basic clinical characteristics were closely related to sarcopenia, and a sarcopenia diagnostic model was developed to predict the presence or absence of sarcopenia. In addition, an analysis of 966 elderly people in the community of Hubei, China, found that its sarcopenia diagnostic model had better diagnostic performance than the three sarcopenia screening models of SARC-F, Ishii and calf circumference.

[0019] However, although the sarcopenia prediction models reported in many studies at home and abroad have shown good diagnostic properties, most of these models are based on cross-sectional survey data from specific regions or populations, and fail to fully reflect that the onset of sarcopenia is a complex process of long-term cumulative effects of multiple factors. Therefore, these models have great limitations in predicting the risk of sarcopenia over a period of time and achieving quantitative stratification and graded management of the disease based on predicted risks.

[0020] In addition, the current sarcopenia prediction model faces many challenges, such as the high risk of bias due to the lack of independent external validation, which limits the confirmation of the true predictive value of the model; if the treatment of continuous variables and the filling method of missing values ​​are not appropriate, it may cause deviations in model fitting; in the variable screening stage, over-reliance on the analysis results of univariate statistical significance may ignore those predictors that are important only in multivariate analysis. These limitations suggest that more rigorous methodology needs to be adopted in future studies to improve the accuracy and generalization ability of the model.

[0021] In view of this, the present disclosure provides a method for predicting the risk of sarcopenia in the elderly. Figure 1 The present invention is a flowchart of a method for predicting the risk of sarcopenia in the elderly according to an embodiment. The method for predicting the risk of sarcopenia in the elderly comprises the following steps: In step S11, sampling indicators of the elderly sarcopenia risk prediction object in multiple dimensions are obtained, and the dimensions of the sampling indicators are obtained according to the prediction sampling indicators; Among them, sampling indicators can be various data or parameters used to measure or describe the health status of the object to be predicted, which can reflect the health status of the object to be predicted in a specific dimension. Dimensions are used to classify or organize different aspects or fields of sampling indicators, such as physiological dimensions (such as height, weight, muscle mass, etc.), biochemical dimensions (such as blood biochemical indicators, hormone levels, etc.), lifestyle dimensions (such as eating habits, exercise habits, etc.), etc. The submission of prediction sampling indicators is obtained by sorting, summarizing and integrating sampling indicators from different sources and dimensions.

[0022] In the disclosed embodiment, sampling indicators that are closely related to the risk of sarcopenia in the elderly are determined. These indicators may include physiological indicators (such as height, weight, muscle mass, grip strength, etc.), biochemical indicators (such as blood biochemical data, hormone levels, etc.), lifestyle indicators (such as eating habits, exercise habits, smoking and drinking conditions, etc.) and genetic indicators (such as the variation of specific genes, etc.). The division of dimensions should follow the following principles: Correlation principle: The sampling indicators within the same dimension should have a high correlation and can jointly reflect a certain aspect of health status. Independence principle: The sampling indicators between different dimensions should be kept as independent as possible to avoid too much overlap or redundant information. Integrity principle: The divided dimensions should be able to comprehensively cover all aspects related to the risk of sarcopenia in the elderly to ensure the integrity of the information.

[0023] In the disclosed embodiment, more than 50 potential predictive sampling indicators were initially screened and classified into five dimensions: demographic characteristics, health-related risk factors, chronic disease-related risk factors, socioeconomic status-related risk factors, and human tissue detection indicators. The focus was on variables that frequently appeared in multiple models (such as age, BMI, and calf circumference) and used them as core building elements.

[0024] In the disclosed embodiment, based on existing research or professional knowledge, sampling indicators and their corresponding dimensions related to the risk of sarcopenia in the elderly are determined. Then, through questionnaire surveys, community physical examinations, laboratory tests, etc., sampling indicators of the object to be predicted in these dimensions are obtained. These sampling indicators will be used for subsequent model input and prediction.

[0025] In step S12, the sampling index corresponding to each dimension is input into the corresponding sub-model in the sarcopenia occurrence risk prediction model to obtain the predicted sarcopenia occurrence result corresponding to the risk prediction object output by the sarcopenia occurrence risk prediction model, wherein the sarcopenia occurrence risk prediction model screens the target features of multiple sub-models according to the corresponding sampling indexes, and then inputs the target features into the multi-factor prediction model results for comprehensive analysis to obtain the predicted sarcopenia occurrence result; Among them, the sub-model is a prediction model built for a specific dimension or a specific sampling index, which is used to extract the target feature in the dimension or the sampling index. The target feature is a sampling index or feature that is identified in the sub-model as being closely related to the risk of sarcopenia. The multi-factor prediction model is a prediction model that can comprehensively analyze the target features extracted by multiple sub-models to obtain the final prediction result.

[0026] In the multivariate logistic regression analysis stage, 17 variables with a P value less than 0.2 in the univariate analysis and 6 variables with significant clinical significance were comprehensively considered. The forward and backward stepwise regression methods were used for the analysis. The results showed that age, gender, BMI, decreased physical activity, malnutrition, pain and calf circumference were independent risk factors for sarcopenia. Among them, the adjusted odds ratio (Adjusted Odds Ratio, Adjusted OR) was used to measure the strength of the association between each independent risk factor and the outcome after controlling the influence of other confounding variables. The 95%CI (Confidence Interval) was used to observe the results in the experiment.

[0027] In the disclosed embodiment, the sampling index corresponding to each dimension is input into the corresponding sub-model. These sub-models have been trained to identify the target features in their respective dimensions. Then, the target features extracted by each sub-model are input into the multi-factor prediction model for comprehensive analysis. The multi-factor prediction model obtains the final prediction result of sarcopenia by considering the interaction and correlation between the target features in different dimensions.

[0028] Among them, the sarcopenia risk prediction model is obtained by training the sub-models of corresponding dimensions through sample sampling indicators of multiple dimensions, analyzing through a multi-element prediction model and through multiple prediction performance evaluations, and then verifying through a validation set in a sarcopenia cohort database. The variance inflation factor is used to evaluate the multicollinearity between the sample sampling indicators to ensure the stability of the sarcopenia risk prediction model. The Bootstrap repeated sampling method is used to simulate various sample distribution scenarios, and the performance parameters of the sarcopenia risk prediction model are repeatedly calculated, and consistent calibration and discrimination indicators are used for performance evaluation.

[0029] In the disclosed embodiments, the training data set is based on the field survey of the Western Health and Aging Trends Study (WCHAT), which systematically collects questionnaire information, anthropometric data, body fluids and tissue specimens. Follow-up survey (July to December 2019): A consistent protocol with the baseline was used to obtain comprehensive follow-up information.

[0030] The external validation data set comes from the established "Chinese Sarcopenia Cohort Database". It was collected through the following stages: 1) Baseline survey (September to December 2021): Establish a database and sample library through on-site surveys to obtain questionnaires and physical examination information from research subjects. 2) Follow-up survey (October 2022 to February 2023): Follow the baseline survey process to enrich follow-up data.

[0031] Among them, difference inflation factor (VIF) is an indicator used to evaluate the multicollinearity between sample sampling indicators. A high VIF value indicates that there is serious multicollinearity between sampling indicators, which may lead to model instability.

[0032] Bootstrap repeated sampling method: A statistical method that simulates different sample distribution scenarios by randomly selecting multiple samples (called Bootstrap samples) with replacement from the original sample to evaluate the stability and performance of the model.

[0033] Calibration and discrimination indicators: indicators used to evaluate the performance of prediction models. Calibration indicators reflect the degree of consistency between the model's prediction results and the actual results, and discrimination indicators reflect the model's ability to distinguish between different risk levels.

[0034] In the disclosed embodiment, the construction process of the sarcopenia risk prediction model includes multiple steps: first, the sample sampling indicators of multiple dimensions are used to train the sub-models of the corresponding dimensions; then, the target features extracted by the sub-models are comprehensively analyzed through the multi-element prediction model; then, the model is evaluated through multiple prediction performance evaluation indicators (such as accuracy, recall rate, F1 score, etc.); finally, the model is verified using the validation set in the sarcopenia cohort database. During the model construction process, the variance inflation factor is used to evaluate the multicollinearity between the sample sampling indicators to ensure the stability of the model; the Bootstrap repeated sampling method is used to simulate a variety of sample distribution scenarios, repeatedly calculate the performance parameters of the model, and use consistent calibration and discrimination indicators for performance evaluation to ensure the accuracy and reliability of the model.

[0035] Variance inflation factor (VIF) was used to assess multicollinearity between variables to ensure model stability. ROC curve and AUC indicators were used to evaluate model discrimination, and the Youden index was combined to determine the optimal prediction threshold. Calibration curve and Hosmer-Lemeshow goodness-of-fit test were used to assess calibration, and DCA and CIC curves were used to analyze the clinical practicality of the model to ensure that the prediction model had both academic value and clinical application potential.

[0036] The robustness of the model was evaluated through internal and external validation. The internal validation used the Bootstrap repeated sampling method to simulate a variety of sample distribution scenarios and repeatedly calculated the model performance parameters to verify the model stability. The external validation used an independent data set to comprehensively evaluate the applicability of the model in different populations, and used consistent calibration and discrimination indicators for performance evaluation to further verify the generalization ability of the model.

[0037] In summary, by obtaining sampling indicators of the elderly in multiple dimensions, combined with the target feature screening of multiple sub-models and comprehensive analysis of multi-factor prediction models, accurate prediction of the risk of sarcopenia in the elderly was achieved. At the same time, the accuracy and reliability of the model were ensured through rigorous model stability and performance evaluation and validation set verification.

[0038] The above technical scheme can more comprehensively reflect the health status of the elderly by obtaining sampling indicators of the elderly in multiple dimensions, thereby improving the accuracy of the risk prediction of sarcopenia. The sampling indicators corresponding to each dimension are input into the corresponding sub-model, and the sub-model performs target feature screening, which helps to identify key features closely related to the occurrence of sarcopenia and reduce the interference of noise data. The target features of multiple sub-models are output to the multi-factor prediction model for comprehensive analysis, which can comprehensively consider the influencing factors in different dimensions and further improve the accuracy and reliability of the prediction. The use of variance inflation factor to evaluate the multicollinearity between sample sampling indicators helps to ensure the stability of the model; the use of Bootstrap repeated sampling method to simulate a variety of sample distribution scenarios, repeatedly calculate the model performance parameters, and use consistent calibration and discrimination indicators for performance evaluation, can comprehensively evaluate the prediction performance of the model and ensure the accuracy and reliability of the model. Verification by the validation set in the sarcopenia cohort database can further verify the prediction effect of the model and ensure the feasibility of the model in practical applications. In summary, by obtaining the sampling indicators of the elderly in multiple dimensions, combining the target feature screening of multiple sub-models and the comprehensive analysis of the multi-factor prediction model, the risk of sarcopenia in the elderly is accurately predicted. At the same time, through rigorous model stability and performance evaluation and validation set verification, the accuracy and reliability of the model are ensured, thereby accurately predicting the risk of sarcopenia in the elderly, thereby achieving efficient early detection, early screening and early intervention of sarcopenia in the elderly.

[0039] In one possible implementation, see Figure 2 As shown, the sub-model includes: a first sarcopenia prediction sub-model based on blood markers; the first sarcopenia prediction sub-model is trained in the following way: In step S21, based on bagging sampling, the input sample sampling index based on the blood marker is subjected to multiple self-service resampling, wherein each sampling of the sample sampling index generates a different data set, thereby obtaining multiple self-service resampling data sets; Among them, bagging generates multiple different training subsets by performing multiple random samplings with replacement on the original data set to increase the diversity and robustness of the model.

[0040] Bootstrap Resampling randomly extracts samples from the original dataset each time (allowing for repetitions) until a new dataset of the same size as the original dataset is formed.

[0041] In the disclosed embodiment, in order to construct diversified training data, a self-help resampling technique is used to randomly extract data from the sample sampling index based on blood markers multiple times. Each sampling allows the same sample to be selected multiple times, and some samples may not be selected once, thereby generating multiple slightly different data sets (self-help resampling data sets). The purpose of this is to simulate different data distributions and improve the generalization ability of subsequent models.

[0042] In the disclosed embodiment, the MICE package is used to perform multiple interpolation for sample sampling indicators to generate multiple complete data sets to enhance the robustness of the model. The continuous variables are truncated to the 1st and 99th percentiles to avoid the influence of extreme values. The normality of the variables is tested to ensure that the data matches the model assumptions.

[0043] In step S22, multiple random forest models are constructed according to the sample sampling index of the self-resampled data set; Among them, the random forest algorithm combines bagging sampling technology and random attribute selection principles to enhance the model's prediction ability by building a large number of decision trees. It effectively reduces the risk of overfitting. Bag sampling generates multiple independent sample sets, combined with random feature selection strategies, which enhances the stability of the model while improving generalization ability. In addition, the random forest algorithm is less sensitive to missing values ​​and outliers, and can handle complex patterns in data sets and provide variable importance rankings.

[0044] Among them, in order to compare the predictive performance of the Logistic regression model and the random forest model in the external validation set, the Delong method was used to analyze the ROC curves of the two models to detect whether there was a significant difference in their classification performance, and the statistical significance was evaluated by hypothesis testing. At the same time, the radar chart was used to visualize the sensitivity, specificity, positive predictive value, negative predictive value and accuracy, and the performance difference between the two models in the independent validation set was intuitively displayed, providing a comprehensive perspective for evaluating their clinical application value. The final Logistic regression model was visualized as a nomogram using the "rms" and "nomogram" packages of R software to display the various predictive factors and their contributions of the sarcopenia risk prediction model. At the same time, an online risk calculation tool was developed using the "DynNom" and "shiny" packages. Users can quickly calculate the probability of sarcopenia by entering the numerical values ​​of the risk factors, and present the results in graphics and data summary windows. This tool provides intuitive and convenient support for clinical applications, making it easier for medical staff to understand the contribution of predictive factors to the results, thereby optimizing decision making.

[0045] In the disclosed embodiments, random forest models are constructed separately. Each random forest model consists of multiple decision trees, and each tree uses only some features and some data during training, which further increases the diversity and robustness of the model. This strategy helps reduce the risk of overfitting of the model and improves the consistency of the model's performance on different data sets.

[0046] In step S23, for each of the random forest models, a variable importance index is used to calculate a feature score of each of the sample sampling indicators in the random forest model, wherein the feature score is used to indicate the importance of the sample sampling indicator in the random forest model; Among them, variable importance measure: used to evaluate the contribution of features in model prediction, usually calculated based on indicators such as the frequency of feature usage in tree splitting or information gain.

[0047] Among them, the random forest algorithm is used to evaluate the impact of 35 variables on the occurrence of sarcopenia, and features are selected according to the importance of the variables. The model hyperparameters, such as the number and maximum depth of decision trees, are optimized by cross-validation to ensure a balance between prediction accuracy and model complexity. The prediction model based on the random forest algorithm is also evaluated by indicators such as ROC curve, AUC and DCA. The variable importance analysis of the model further reveals the key predictive factors. Combined with visualization tools such as feature importance heat maps, the results are more intuitive and valuable to interpret.

[0048] In the disclosed embodiment, for each random forest model, a variable importance index is used to quantify the contribution of each sample sampling index in the model, i.e., a feature score. This score reflects the direct influence of the feature on the model prediction results, and high-scoring features have a more significant impact on the prediction results.

[0049] In step S24, the sample sampling indicators are disrupted in order, noise is added and randomized to generate shadow features, and the shadow features are marked with feature scores; Among them, Shadow Feature is a virtual feature generated by randomization or adding noise, which is used to evaluate the robustness of feature importance.

[0050] In the disclosed embodiment, in order to verify the stability and reliability of feature scoring, the original sample sampling indicators are shuffled, noise is added or randomized to generate a series of shadow features. These shadow features retain the structure of the original data but destroy the true relationship between them and the target variable. These shadow features are then assigned feature scores to compare their differences with the true feature scores.

[0051] In step S25, an iterative calculation is performed based on the feature score of the sample sampling index in the random forest model and the feature score of the corresponding factor feature until the preset training conditions are met to obtain the first sarcopenia prediction sub-model.

[0052] In the disclosed embodiment, the feature scores of the real features and the feature scores of the shadow features are combined to continuously optimize the model parameters through iterative calculation. During the iteration process, the feature selection may be adjusted according to the importance of the features, unimportant features may be eliminated, or the parameters of the random forest (such as the number of trees, maximum depth, etc.) may be adjusted. This process continues until the preset training conditions are met, such as when the model performance is stable or a predetermined number of iterations is reached. Finally, the obtained model is the first sarcopenia prediction sub-model, which can effectively predict the risk of sarcopenia based on blood markers.

[0053] In a possible implementation, in step S25, the feature score of the random forest model and the feature score of the corresponding factor feature are iteratively calculated according to the sample sampling index until the preset training condition is reached to obtain the first sarcopenia prediction sub-model, including: In step S251, according to the feature score of the sample sampling indicator in the random forest model and the feature score of the corresponding factor feature, for any of the sample sampling indicators, the sample sampling indicator and the feature score of the corresponding shadow feature are compared to determine a relatively important feature, wherein if the feature score of the sample sampling indicator is higher than the feature score of the corresponding shadow feature, the sample sampling indicator is a relatively important feature; Among them, Relatively Important Features: Among all the features, those that show higher feature scores relative to their shadow features are considered to be more critical features in the model.

[0054] In the disclosed embodiment, this is achieved by comparing the feature score of each sample sampling indicator with the feature score of its corresponding shadow feature. Shadow features are generated by randomization, and they are similar in form to the original features, but in theory should not contain any information related to the target variable (here, the risk of sarcopenia). Therefore, if the feature score of a sample sampling indicator is significantly higher than the feature score of its shadow feature, this indicates that the sample sampling indicator is likely to contain information related to the target variable and is therefore considered to be a relatively important feature.

[0055] In step S252, based on the relatively important features, multiple self-service resampling of the relatively important features is performed to regenerate a random forest and generate the corresponding shadow features, and the feature scores of the relatively important features and the corresponding shadow features determined in the previous round are used for iterative calculation until the preset training conditions are met to obtain the first sarcopenia prediction sub-model.

[0056] Among them, iterative resampling is to repeatedly sample data according to certain rules during the model training process to continuously optimize the model performance.

[0057] Bootstrapped resampling: Perform multiple bootstrapped resampling only for the relatively important features determined in the previous step. This is done to focus on those features that have been proven to contribute substantially to the model's predictive performance while reducing the impact of noise and redundant information.

[0058] Model reconstruction: Use the resampled dataset to build a new random forest model. This step is to verify and possibly improve the performance of the model on the new data distribution.

[0059] Shadow feature generation and iterative calculation: For each rebuilt random forest model in each iteration, new shadow features need to be generated for its relatively important features and their feature scores calculated. These feature scores are then used to update the understanding of feature importance and provide information for the next round of iterations.

[0060] Preset training condition judgment: At the end of each round of iteration, it will be checked whether the preset training conditions are met. These conditions may include the prediction accuracy of the model, the upper limit of the number of iterations, the stability of the feature score, etc. Only when these conditions are met will the iteration process stop, and the model obtained at this time is the final first sarcopenia prediction sub-model.

[0061] Through this process, the first sarcopenia prediction sub-model is able to gradually focus on those features that contribute substantially to the prediction performance, while reducing the impact of noise and redundant information, thereby improving the accuracy and robustness of the model.

[0062] In a possible implementation, in step S23, for each of the random forest models, using the variable importance index to calculate the feature score of each of the sample sampling indicators in the random forest model includes: In step S231, for each of the random forest models, the Gini index is calculated at each node of the decision tree; Among them, the Gini Index is an indicator used to measure the impurity of data and is used to select split nodes in the decision tree algorithm. The closer the value of the Gini Index is to 0, the lower the impurity of the data.

[0063] In the disclosed embodiment, for each decision tree in the random forest model, the Gini index needs to be calculated at each node. The calculation of the Gini index is based on the proportion of samples of each category in the node. Assuming that there are k categories in node T and the proportion of samples of the i-th category is p_i, the Gini index Gini(T) of node T can be expressed as: Gini(T) = 1 - Σ(p_i^2), where Σ represents the sum over all categories.

[0064] In step S232, the amount of impurity reduction each time the sample sampling index is used to split in the decision tree is determined by the difference in the Gini index before and after the split; Among them, the impurity reduction is the reduction in node impurity caused by selecting a feature for splitting during the decision tree splitting process.

[0065] In the disclosed embodiment, for each split in each decision tree, it is necessary to calculate the Gini index difference before and after the split, that is, the amount of impurity reduction. Specifically, assuming that the Gini index of node T before the split is Gini(T), the two child nodes obtained after the split are T_left and T_right, their Gini indexes are Gini(T_left) and Gini(T_right), and the proportion of nodes after the split to the total samples is w_left and w_right, respectively. Then the amount of impurity reduction ΔGini caused by the split can be expressed as: ΔGini = Gini(T) - (w_left * Gini(T_left) + w_right * Gini(T_right)) In step S233, the average impurity reduction in all the decision trees is determined based on the impurity reduction of each sample sampling indicator on all the decision trees and the number of decision trees, and the average impurity reduction is used as the feature score of the sample sampling indicator.

[0066] Among them, Average Impurity Reduction: The average value of the impurity reduction caused by a sample sampling indicator in all decision trees is used as the feature score of the sample sampling indicator.

[0067] In the disclosed embodiment, for each sample sampling indicator, it is necessary to calculate its average impurity reduction in all decision trees. This can be achieved by accumulating the impurity reduction caused by the sample sampling indicator in each decision tree and then dividing it by the total number of decision trees. The average value obtained is the feature score of the sample sampling indicator. The higher the feature score, the greater the contribution of the sample sampling indicator in reducing data impurity and improving the prediction ability of the model.

[0068] Through this process, a quantitative feature score can be assigned to each sample sampling indicator, which reflects the importance of the indicator in the random forest model. These feature scores will be used in subsequent feature selection, model optimization and other steps.

[0069] In one possible implementation, see Figure 2 As shown, the sub-model includes: a second sarcopenia prediction sub-model based on intestinal flora; the second sarcopenia prediction sub-model is trained in the following way: In step S31, PCR amplification is performed on the conserved region in the 16S rRNA gene using universal primers to obtain amplified gene fragments, and the amplified gene fragments are added to sequencing adapters to construct a 16S rRNA gene library; Among them, the 16S rRNA gene: is a subunit of bacterial ribosomal RNA, its sequence is highly conserved and interspecies different, and is often used for bacterial classification and identification. PCR amplification (Polymerase Chain Reaction Amplification): is a molecular biology technique used to quickly replicate specific DNA fragments in vitro. Sequencing Adaptors: are used to connect short DNA sequences between PCR amplification products and high-throughput sequencing platforms to help sequencers identify and read sample information.

[0070] In the disclosed embodiment, the conserved regions of the 16S rRNA gene in the sample are PCR amplified using universal primers in order to increase the number of target DNA fragments. The amplified gene fragments are added to sequencing adapters, which act as "bridges" to connect the amplified products to the sequencing machine. Through this step, a library containing all bacterial 16S rRNA gene fragments in the sample is constructed.

[0071] In step S32, the 16S rRNA gene library is sequenced using high-throughput sequencing technology to obtain 16S rRNA gene sequence data, and the 16S rRNA gene sequence obtained by sequencing is annotated with species to determine the types of microorganisms in the sample; Among them, high-throughput sequencing: is a technology that can sequence millions of DNA fragments at the same time, greatly improving the sequencing speed and efficiency. Species Annotation: Identify and annotate the types of microorganisms present in the sample based on the DNA sequence information obtained by sequencing.

[0072] In the disclosed embodiment, the 16S rRNA gene library constructed in step S31 is sequenced using high-throughput sequencing technology to obtain a large amount of gene sequence data. These sequence data are analyzed by bioinformatics, such as comparison databases and algorithm predictions, to identify the various microbial species present in the sample, i.e., species annotation. This step provides detailed information on the microbial community structure in the sample.

[0073] In step S33, the abundance of sample populations of different age groups in the samples is counted to obtain the distribution of sample populations of different age groups in sarcopenia patients and healthy controls, and based on the distribution, the microbial community structure and differences of the sample populations of different age groups are analyzed through α diversity and β diversity; Among them, Alpha Diversity: an indicator to measure the richness and uniformity of a sample population. Beta Diversity: an indicator to measure the composition differences between different sample populations.

[0074] In the disclosed embodiments, the abundance of microbial species in sarcopenia patients and healthy controls in sample populations of various age groups was statistically analyzed, that is, the number or proportion of each microorganism in the sample. Then, α diversity and β diversity analysis were used to evaluate the richness, uniformity, and composition differences of microbial communities in different age groups and different health states. These analyses help reveal the relationship between microbial community structure and sarcopenia.

[0075] In step S34, a multivariate linear model is used to analyze the microbial community structure and differences to identify characteristic intestinal flora significantly associated with the occurrence of sarcopenia, and obtain flora analysis results; Among them, Multivariate Linear Model: A statistical model used to study the relationship between multiple independent variables and dependent variables.

[0076] In the disclosed embodiment, a multivariate linear model is used to analyze the microbial community structure and differences obtained in step S33. This model can help us identify which microbial community characteristics (such as the abundance of a specific flora) are significantly associated with the occurrence of sarcopenia. Through this step, the flora analysis results can be obtained, that is, a series of characteristic intestinal flora significantly associated with sarcopenia.

[0077] In step S35, based on the results of the flora analysis, sample intestinal flora that is significantly correlated with the occurrence of sarcopenia is screened out, and among the sample intestinal flora, the characteristics corresponding to the target sample intestinal flora that are representative and whose prediction performance meets the preset requirements are selected as input variables of the model to construct a prediction model, thereby obtaining the second sarcopenia prediction sub-model.

[0078] In the disclosed embodiment, according to the flora analysis results obtained in step S34, the intestinal flora of the samples significantly associated with the occurrence of sarcopenia is screened out. Then, among these significantly associated intestinal flora, the characteristics corresponding to the intestinal flora of the target sample that are representative and have good predictive performance are selected as the input variables of the model. These characteristics may be the abundance, proportion, or combination of certain specific flora. Finally, a prediction model is constructed using a machine learning algorithm (such as logistic regression, support vector machine, random forest, etc.), and a second sarcopenia prediction sub-model is trained. This model can predict the risk of an individual suffering from sarcopenia based on the intestinal flora characteristics of the sample.

[0079] Through this series of steps, a second sarcopenia prediction sub-model based on intestinal flora was trained, which provides new tools and methods for the early detection and prevention of sarcopenia.

[0080] In one possible implementation, see Figure 2 As shown, the sub-model includes: a third sarcopenia prediction sub-model based on clinical characteristics; the third sarcopenia prediction sub-model is trained in the following way: In step S41, a univariate logistic regression analysis is used to preliminarily evaluate the correlation between the variables and the occurrence of sarcopenia, wherein the sample sampling indicators and the labeled sample sampling indicators with probability values ​​less than a preset threshold are included in the multivariate analysis, wherein the labeled sample sampling indicators are used to identify the clinical limitations of the sample sampling indicators; Among them, Single-Factor Logistic Regression Analysis is used to evaluate the strength of association between a single independent variable and a dependent variable (usually a binary variable, such as whether or not the patient is ill). Probability value (P-value): In statistics, an indicator used to measure the significance of the difference between the observed data and the null hypothesis. Generally, a smaller P value indicates that the difference between the data and the null hypothesis is more significant. Predefined Threshold: In statistical analysis, a probability value standard is set in advance to determine which variables are significant. Labeled Sample Variables refers to sample characteristics that are specially labeled because of their clinical importance or limitations in addition to being statistically significant.

[0081] Univariate logistic regression analysis was used to evaluate the correlation between each sample sampling indicator (such as age, gender, biochemical indicators, etc.) and the occurrence of sarcopenia. The purpose of this step is to preliminarily screen out those variables that are significantly associated with the occurrence of sarcopenia. Specifically, the probability value (P-value) of each variable is calculated, and those variables with probability values ​​less than the preset threshold and those variables that are specially marked due to clinical importance or restriction are included in the subsequent multivariate analysis. Such a screening process helps to reduce noise variables and improve the accuracy and interpretability of subsequent models.

[0082] A sarcopenia risk prediction model was constructed by Logistic regression model to screen potential independent risk factors. Univariate Logistic regression analysis was used to preliminarily evaluate the correlation between variables and the occurrence of sarcopenia. Variables with P values ​​< 0.2 and clinically significant variables were included in multivariate analysis, and forward and backward stepwise regression method was used to optimize variable selection. Finally, the optimal model was selected based on the Akaike information criterion (AIC) value, and the odds ratio (OR) and its 95% confidence interval (CI) were calculated to evaluate the independence of variables.

[0083] In step S42, the sample sampling index selection is optimized using the forward and backward stepwise regression method to obtain the screening sampling index; Among them, the forward-backward stepwise regression method optimizes the model by gradually adding or deleting variables. Forward stepwise regression starts from an empty model and gradually adds variables; backward stepwise regression starts from the full model and gradually deletes variables. Forward-backward stepwise regression combines the advantages of both methods, and can consider deleting variables that are no longer significant while adding new variables.

[0084] In the disclosed embodiment, the variables screened in step S41 are further optimized using the forward and backward stepwise regression method. The purpose of this step is to find an optimal combination of variables so that the model is both concise and accurate. Specifically, the final combination of variables is determined by gradually adding or deleting variables and evaluating the performance of the model after each change (such as the Akaike Information Criterion value). The Akaike Information Criterion (AIC) is an indicator for measuring the goodness of fit of a model. It takes into account the complexity of the model and the fit of the data. The model with the smallest AIC value is usually selected as the optimal model.

[0085] In step S43, model training is performed according to the screening sampling indicators, and the optimal model is selected based on the Akaike information criterion value, the odds ratio is calculated, and the screening sampling indicators whose odds ratio is within the preset confidence interval of the optimal model are used as the target features output by the third sarcopenia prediction sub-model.

[0086] Among them, Akaike Information Criterion (AIC) is used as a criterion for model selection, which comprehensively considers the model's fit and complexity. The smaller the AIC value, the better the model. Odds Ratio (OR): In a logistic regression model, it indicates the change in the ratio of the probability of the dependent variable occurring for each unit increase in the independent variable. Predefined Confidence Interval: In statistical analysis, in order to evaluate the reliability of the estimated value, a pre-set interval range is usually expressed as a 95% confidence interval.

[0087] In the disclosed embodiment, model training is performed according to the screening sampling indicators obtained in step S42. Then, the optimal model is selected based on the Akaike Information Criterion value, that is, the model with the smallest AIC value. Next, the odds ratio (OR) of each screening sampling indicator in the optimal model is calculated, and it is evaluated whether these odds ratios are significant within the preset confidence interval. Finally, the screening sampling indicators corresponding to the odds ratios that are significant within the preset confidence interval of the optimal model are used as the target features of the output of the third sarcopenia prediction sub-model. These feature screening results not only reflect the variables that are significantly related to the occurrence of sarcopenia, but also improve the accuracy and stability of the prediction through the optimization process of the model.

[0088] Through this series of steps, a third sarcopenia prediction sub-model based on multivariate logistic regression analysis was trained. This model can predict the risk of sarcopenia based on an individual's sample sampling indicators and provide important reference for clinical decision-making.

[0089] In a possible implementation, the method includes: In step S51, the probability of occurrence of sarcopenia corresponding to the risk prediction object is calculated by applying the coefficients in the sarcopenia occurrence risk prediction model, wherein a plurality of key prediction variables are integrated in the nomogram model, and the value of each key prediction variable is represented by a vertical line and intersects with a preset scoring scale in the nomogram; Among them, the nomogram model: a visualization tool used to convert complex statistical prediction models into intuitive graphical representations. In the nomogram, each predictor variable (such as age, gender, biochemical indicators, etc.) has a corresponding vertical line, through which the contribution value or score of the variable in the model can be read. Coefficient: In a statistical model, a numerical value that indicates the degree of influence of an independent variable on a dependent variable. In a logistic regression model, the coefficient determines how the logarithm of the dependent variable (here, the probability of sarcopenia) changes when the independent variable increases by one unit.

[0090] In the disclosed embodiment, the coefficients in the trained sarcopenia risk prediction model are applied. These coefficients reflect the degree of influence of each key predictive variable on the probability of sarcopenia. Then, in the nomogram model, the numerical value of each key predictive variable of the risk prediction object is converted into a point on the vertical line and intersects with the preset scoring scale in the nomogram. This scoring scale is pre-set based on the model coefficients and the value range of the variable, and is used to convert the numerical value of the variable into a score. Through this step, the score of the risk prediction object on each key predictive variable can be obtained.

[0091] In step S52, the scores of the risk prediction object on the scoring scale are accumulated for the multiple key prediction variables to obtain a total score of the risk prediction object; In the disclosed embodiment, the scores of each key prediction variable obtained in step S51 are accumulated to obtain a total score of the risk prediction object. This total score reflects the comprehensive performance of the risk prediction object on all key prediction variables and is an important basis for predicting the probability of sarcopenia.

[0092] In step S53, the corresponding score is found on the total score axis of the nomogram, and a vertical line is drawn downward from the point where the corresponding score is located on the total score axis, and the value corresponding to the intersection of the vertical line and the risk prediction value axis is determined as the probability of occurrence of sarcopenia in the risk prediction subject within a preset time length; In the disclosed embodiment, first find the score point corresponding to the total score of the risk prediction object on the total score axis of the nomogram. Then, draw a vertical line downward from this point until it intersects with the risk prediction value axis. The value corresponding to this intersection is the probability of sarcopenia in the risk prediction object within a preset time length (such as one year, five years, etc.). This probability value is obtained through model prediction, which reflects the possibility of the risk prediction object suffering from sarcopenia in a certain period of time in the future.

[0093] In step S54, the predicted sarcopenia occurrence result of the risk prediction object is determined according to the sarcopenia occurrence probability of the risk prediction object.

[0094] In the disclosed embodiment, the predicted sarcopenia occurrence result of the risk prediction object is determined according to the probability of sarcopenia occurrence obtained in step S53. This is usually achieved by setting a probability threshold. For example, if the set threshold is 0.5 (i.e., a 50% probability), then when the probability of sarcopenia occurrence of the risk prediction object is greater than or equal to 0.5, it can be considered that it has a higher risk of developing sarcopenia in a certain period of time in the future; otherwise, it is considered that the risk is low. This prediction result can provide an important reference for clinical decision-making and help doctors develop personalized prevention and treatment strategies.

[0095] Through this series of steps, the nomogram model is used to transform the complex sarcopenia risk prediction model into an intuitive graphical representation, and the probability of sarcopenia occurrence is calculated based on the actual data of the risk prediction object.

[0096] In a possible implementation, the regression formula of the sarcopenia risk prediction model is: P=1 / [1+exp^-(a+bx+m×ng×h+c×d+j×k+L×z+q×w)]; Among them, P is the probability of occurrence of sarcopenia in the risk prediction object within a preset time length, a is the prediction model parameter, b is the age factor, x is the age of the risk prediction object, m is the gender factor, n is the gender, g is the body mass index factor, h is the body mass index of the risk prediction object, c is the physical activity factor, d is the physical activity parameter of the risk prediction object, j is the nutritional factor, k is the nutritional status parameter of the risk prediction object, L is the pain factor, z is the pain parameter of the risk prediction object, q is the leg circumference factor, and w is the calf circumference of the risk prediction object.

[0097] In the disclosed embodiment, a PRE-SARC (PREdiction of SARCopenia Risk) model is used to predict the risk of sarcopenia in elderly people in the community within 1 year. The model calculates the probability of sarcopenia for each subject by applying the coefficients in the model. The regression formula of the PRE-SARC sarcopenia risk prediction model is: Risk of sarcopenia within 1 year (%) = 1 / (1+exp^[-(12.35+0×elderly people aged 60 to 69 or 1.07×elderly people aged 70 to 79 or 1.54×elderly people aged ≥80+0×female or 0.62×male-0.23×BMI+0.84×low physical activity+1.00×malnutrition+0.52×pain-0.30×calf circumference]).

[0098] With the help of the rms package in R software, the PRE-SARC model was used to visualize the probability of sarcopenia in patients as a nomogram model, which integrated 7 key predictive variables, including age, sex, BMI (body mass index), decreased physical activity, malnutrition, pain, and calf circumference. In the nomogram, the specific value of each variable is represented by a vertical line and intersects with the "score scale" at the top of the figure to obtain the corresponding score of the variable. The patient's total score is calculated by adding up all the variable scores, and then the corresponding score is found on the "total score" axis of the nomogram, and a vertical line is drawn downward from the corresponding score point on the total score axis. The intersection of this line with the "risk prediction value" axis represents the patient's risk prediction probability of sarcopenia within one year. The nomogram of the PRE-SARC model shows the total score range from 0 to 280 points, and the corresponding sarcopenia risk probability, which varies between 0.1 and 0.9.

[0099] Taking a specific case as an example, when a female patient aged 70 to 79 has a BMI of 22, has decreased physical activity, malnutrition and pain symptoms, and a calf circumference of 35.6 cm, the user only needs to enter these indicators into the calculator's operating interface, and the data summary window will show that the patient's risk of sarcopenia is 51.1%. Such results indicate that the patient has a high risk of sarcopenia. This online risk calculator, with its intuitive and simple design, allows clinical medical staff to easily calculate the individual's risk probability of sarcopenia. It helps medical professionals to promptly identify high-risk groups for sarcopenia, thereby promoting early prevention and intervention measures and achieving precise management of geriatric diseases.

[0100] After verification, such as Figure 3As shown in Figure 2, the model's ability to distinguish was evaluated by the area under the ROC curve (AUC), and the results showed that the model's AUC was 0.87 (95% CI, 0.83-0.90). This indicates that the prediction model performs well in distinguishing between patients with sarcopenia and non-sarcopenia (see Figure 3 The optimal prediction probability cutoff value corresponding to the maximum Youden index was 9%. At this cutoff value, the sensitivity of the model was 79.3% (95% CI: 69.0%-87.0%) and the specificity was 80.3% (95% CI: 77.6%-82.8%).

[0101] The present disclosure also provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps of the method described in any one of the aforementioned embodiments are implemented.

[0102] The present disclosure also provides an electronic device, including: a memory having a computer program stored thereon; A processor is used to execute the computer program in the memory to implement the steps of the method in any one of the aforementioned embodiments.

[0103] Figure 4 The device 100 for predicting the risk of sarcopenia in the elderly shown includes: a processor 1001 and a memory 1003. The processor 1001 and the memory 1003 are connected, such as through a bus 1002. Optionally, the device 100 for predicting the risk of sarcopenia in the elderly may also include a communication component 1004, which may be used for data interaction between the device 100 and other devices, such as data transmission and / or data reception. It should be noted that in actual scheduling, the communication component 1004 is not limited to one, and the structure of the device 100 for predicting the risk of sarcopenia in the elderly does not constitute a limitation on the embodiments of the present application.

[0104] Processor 1001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. It may implement or execute various exemplary logic blocks, modules and circuits described in conjunction with the disclosure of this application. Processor 1001 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0105] The bus 1002 may include a path to transmit information between the above components. The bus 1002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus 1002 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 4 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0106] The memory 1003 may be a ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compressed optical disk, laser disk, optical disk, digital versatile disk, Blu-ray disk, etc.), magnetic disk storage medium, other magnetic storage devices, or any other medium that can be used to carry or store program code and can be read by a computer, without limitation herein.

[0107] The memory 1003 is used to store program codes for executing the embodiments of the present disclosure, and the execution is controlled by the processor 1001. The processor 1001 is used to execute the program codes stored in the memory 1003 to implement the steps shown in the embodiment of the method for predicting the risk of sarcopenia in the elderly.

[0108] The embodiments of the present disclosure also provide a computer-readable storage medium having program code stored thereon. When the program code is executed by a processor, the steps and corresponding contents of the aforementioned method embodiment for predicting the risk of sarcopenia in the elderly can be implemented.

[0109] The preferred embodiments of the present disclosure are described in detail above in conjunction with the accompanying drawings; however, the present disclosure is not limited to the specific details in the above embodiments; within the technical concept of the present disclosure, various changes, modifications, substitutions and variations may be made to these embodiments, and these changes, modifications, substitutions and variations all fall within the protection scope of the present disclosure.

[0110] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any appropriate manner without contradiction, and they should also be regarded as the contents disclosed in this disclosure. In order to avoid unnecessary repetition, this disclosure will not further describe various possible combinations. The technical scope of this application is not limited to the contents in the specification, and its technical scope must be determined according to the scope of the claims.

Claims

1. A method for predicting the risk of sarcopenia in the elderly, characterized in that: The method comprises: Acquire sampling indicators in multiple dimensions for the subject to be predicted for the risk of sarcopenia in the elderly, wherein the dimensions of the sampling indicators are obtained by summarizing the predicted sampling indicators; Inputting the sampling index corresponding to each dimension into the corresponding sub-model in the sarcopenia occurrence risk prediction model to obtain the predicted sarcopenia occurrence result corresponding to the risk prediction object output by the sarcopenia occurrence risk prediction model, wherein the sarcopenia occurrence risk prediction model screens the target features of multiple sub-models according to the corresponding sampling indexes, and then inputs the target features into the multi-factor prediction model results for comprehensive analysis to obtain the predicted sarcopenia occurrence result; Among them, the sarcopenia risk prediction model is obtained by training the sub-models of corresponding dimensions through sample sampling indicators of multiple dimensions, analyzing through a multi-element prediction model and through multiple prediction performance evaluations, and then verifying through a validation set in a sarcopenia cohort database. The variance inflation factor is used to evaluate the multicollinearity between the sample sampling indicators to ensure the stability of the sarcopenia risk prediction model. The Bootstrap repeated sampling method is used to simulate various sample distribution scenarios, and the performance parameters of the sarcopenia risk prediction model are repeatedly calculated, and consistent calibration and discrimination indicators are used for performance evaluation.

2. The method for predicting the risk of sarcopenia in the elderly according to claim 1, characterized in that: The sub-model includes: a first sarcopenia prediction sub-model based on blood markers; the first sarcopenia prediction sub-model is trained in the following manner: Based on bagged sampling, the input sample sampling index based on the blood marker is subjected to multiple self-service resampling, wherein each sampling of the sample sampling index generates a different data set, thereby obtaining multiple self-service resampling data sets; Constructing multiple random forest models according to the sample sampling index of the self-resampled data set; For each of the random forest models, a variable importance index is used to calculate a feature score of each of the sample sampling indicators in the random forest model, wherein the feature score is used to indicate the importance of the sample sampling indicator in the random forest model; The sample sampling indicators are disrupted in order, noise is added and randomized to generate shadow features, and the shadow features are marked with feature scores; Iterative calculation is performed according to the feature score of the sample sampling index in the random forest model and the feature score of the corresponding factor feature until the preset training conditions are met to obtain the first sarcopenia prediction sub-model.

3. The method for predicting the risk of sarcopenia in the elderly according to claim 2, characterized in that: The iterative calculation of the feature scores of the random forest model and the feature scores of the corresponding factor features according to the sample sampling index until the preset training conditions are met to obtain the first sarcopenia prediction sub-model includes: According to the feature score of the sample sampling indicator in the random forest model and the feature score of the corresponding factor feature, for any of the sample sampling indicators, the sample sampling indicator and the feature score of the corresponding shadow feature are compared to determine a relatively important feature, wherein if the feature score of the sample sampling indicator is higher than the feature score of the corresponding shadow feature, the sample sampling indicator is a relatively important feature; According to the relatively important features, multiple self-service resampling of the relatively important features is performed to regenerate the random forest and generate the corresponding shadow features, and the feature scores of the relatively important features and the corresponding shadow features determined in the previous round are used for iterative calculation until the preset training conditions are met to obtain the first sarcopenia prediction sub-model.

4. The method for predicting the risk of sarcopenia in the elderly according to claim 3, characterized in that: For each of the random forest models, using the variable importance index to calculate the feature score of each of the sample sampling indicators in the random forest model includes: For each of the random forest models, calculating the Gini index at each node of the decision tree; Determine the reduction in impurity each time the sample sampling index is used to split in the decision tree by using the difference in the Gini index before and after the split; According to the impurity reduction of each sample sampling indicator on all the decision trees and the number of decision trees, the average impurity reduction in all the decision trees is determined, and the average impurity reduction is used as the feature score of the sample sampling indicator.

5. The method for predicting the risk of sarcopenia in the elderly according to claim 1, characterized in that: The sub-model includes: a second sarcopenia prediction sub-model based on intestinal flora; the second sarcopenia prediction sub-model is trained in the following way: Using universal primers to perform PCR amplification on the conserved region in the 16S rRNA gene to obtain an amplified gene fragment, and adding the amplified gene fragment to a sequencing adapter to construct a 16S rRNA gene library; Sequencing the 16S rRNA gene library using high-throughput sequencing technology to obtain 16S rRNA gene sequence data, and annotating the 16S rRNA gene sequence obtained by sequencing to determine the species of microorganisms in the sample; Counting the abundance of sample populations of different age groups in the samples, obtaining the distribution of sample populations of different age groups in sarcopenia patients and healthy controls, and analyzing the microbial community structure and differences of the sample populations of different age groups through α diversity and β diversity based on the distribution; The microbial community structure and differences are analyzed using a multivariate linear model to identify characteristic intestinal flora significantly associated with the occurrence of sarcopenia and obtain flora analysis results; According to the results of the microbiota analysis, the sample intestinal flora that is significantly correlated with the occurrence of sarcopenia is screened out, and among the sample intestinal flora, the characteristics corresponding to the target sample intestinal flora that are representative and whose prediction performance meets the preset requirements are selected as input variables of the model to construct a prediction model, thereby obtaining the second sarcopenia prediction sub-model.

6. The method for predicting the risk of sarcopenia in the elderly according to claim 1, characterized in that: The sub-model includes: a third sarcopenia prediction sub-model based on clinical characteristics; the third sarcopenia prediction sub-model is trained in the following manner: A univariate logistic regression analysis was used to preliminarily evaluate the correlation between the variables and the occurrence of sarcopenia, wherein the sample sampling indicators and the labeled sample sampling indicators with probability values ​​less than a preset threshold were included in the multivariate analysis, wherein the labeled sample sampling indicators were used to identify the clinical limitations of the sample sampling indicators; Utilizing the forward and backward stepwise regression method to optimize the sample sampling index selection, and obtain the screening sampling index; Model training is performed according to the screening sampling indicators, and the optimal model is selected based on the Akaike information criterion value, the odds ratio is calculated, and the screening sampling indicators whose odds ratio is within the preset confidence interval of the optimal model are used as the target features output by the third sarcopenia prediction sub-model.

7. The method for predicting the risk of sarcopenia in the elderly according to any one of claims 1 to 6, characterized in that: The method comprises: The probability of occurrence of sarcopenia corresponding to the risk prediction object is calculated by applying the coefficients in the sarcopenia occurrence risk prediction model, wherein a plurality of key prediction variables are integrated in the nomogram model, and the value of each key prediction variable is represented by a vertical line and intersects with a preset scoring scale in the nomogram; Accumulating the scores of the risk prediction object on the scoring scale for multiple key prediction variables to obtain a total score for the risk prediction object; Find the corresponding score on the total score axis of the nomogram, draw a vertical line downward from the point on the total score axis where the corresponding score is located, and determine the value corresponding to the intersection of the vertical line and the risk prediction value axis as the probability of occurrence of sarcopenia in the risk prediction subject within a preset time length; According to the probability of sarcopenia occurrence of the risk prediction object, the predicted sarcopenia occurrence result of the risk prediction object is determined.

8. The method for predicting the risk of sarcopenia in the elderly according to claim 7, characterized in that: The regression formula of the sarcopenia risk prediction model is: P=1 / [1+exp^-(a+bx+m×ng×h+c×d+j×k+L×z+q×w)]; Among them, P is the probability of occurrence of sarcopenia in the risk prediction object within a preset time length, a is the prediction model parameter, b is the age factor, x is the age of the risk prediction object, m is the gender factor, n is the gender, g is the body mass index factor, h is the body mass index of the risk prediction object, c is the physical activity factor, d is the physical activity parameter of the risk prediction object, j is the nutritional factor, k is the nutritional status parameter of the risk prediction object, L is the pain factor, z is the pain parameter of the risk prediction object, q is the leg circumference factor, and w is the calf circumference of the risk prediction object.

9. An electronic device, characterized in that: include: a memory having a computer program stored thereon; A processor, configured to execute the computer program in the memory to implement the steps of the method according to any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method described in any one of claims 1 to 8 are implemented.

Images