Method and system for constructing sarcopenia prediction model without physique measurement

Abstract

The invention relates to a sarcopenia prediction model construction method and system without physique measurement, and belongs to the field of medical informatization. The method comprises the steps of establishing a crowd data set, preprocessing sample data, dividing a training set and a test set, screening modeling features, constructing a plurality of machine learning classifiers, selecting an optimal model, evaluating and predicting performance and the like. The high-performance sarcopenia prediction model can be constructed by only using information such as age, gender, height, weight and the like which can be obtained through questions and answers or self-evaluation as independent variables without physical fitness measurement such as grip strength and physical fitness test. The system comprises a data acquisition module, a data analysis module and a result output module, can realize rapid, low-cost, remote and self-evaluation prediction of sarcopenia risk, is suitable for scenes of various medical institutions, communities, families and the like at all levels, and is beneficial to early screening, intervention and dynamic monitoring of sarcopenia.

Description

Technical Field

[0001] The present invention belongs to the field of medical information technology and relates to a method and system for constructing a sarcopenia prediction model without the need for physical measurement. Background Art

[0002] Sarcopenia, sometimes also called sarcopenia or amniotic fluid, is a pathological condition characterized by age-related loss of skeletal muscle mass, muscle strength and / or physical fitness. Research statistics show that the prevalence of sarcopenia ranges from 5.5% to 25.7%, with a higher incidence in men of about 5.1%-21.0% and in women of about 4.1%-16.3%. However, due to insufficient public health and clinical knowledge of the disease, the actual disease burden of sarcopenia is still seriously underestimated. The global prevalence of sarcopenia is expected to increase from 50 million in 2010 to 200 million in 2050, indicating that sarcopenia has gradually become a major global public health issue that must be effectively addressed and addressed.

[0003] Sarcopenia can occur in community populations without other chronic diseases, or it can occur with other diseases. For this reason, it has received widespread attention from researchers and clinicians from different disciplines. Studies have found that sarcopenia can cause serious physiological and clinical consequences, such as disability, falls, cardiovascular events, impaired quality of life, and increased mortality, but effective treatment measures are very limited. The International Classification of Diseases has assigned a specific disease code M62.84 to sarcopenia in the 10th edition of the Clinical Modification (ICD-10-CM), marking the recognition of sarcopenia as an independent disease entity. It is worth noting that with the continued growth of the aging population, functional disability-related issues represented by sarcopenia will become more prominent. Whether in China or around the world, clinical cognition and research on the screening, diagnosis, treatment and monitoring of sarcopenia will continue to emerge and progress.

[0004] In the past decade, at least six diagnostic frameworks for sarcopenia have been proposed, two of which have been widely recognized in both research and clinical scenarios, including the updated guidelines of the European Working Group on Sarcopenia in Older People (EWGSOP2) for Western populations and the 2019 consensus of the Asian Working Group on Sarcopenia (AWGS). Both are similar in terms of the definition and indicators of sarcopenia, including the evaluation of muscle strength, physical condition, and muscle mass.

[0005] However, the existing technology has the following problems in the complete diagnosis of sarcopenia: 1) individuals need to go to medical institutions in person for treatment; 2) they need to use a handgrip dynamometer to measure grip strength according to standardized technical solutions under the guidance of health professionals; 3) they need to perform physical fitness tests under the supervision and guidance of health professionals, such as completing a 6-meter walking speed, a five-time stand test, or an SPPB scale assessment; 4) they need to use dual-energy X-ray absorptiometry (DEXA) or bioelectrical impedance analysis (BIA) and other equipment to measure the skeletal muscle mass of individuals and calculate the skeletal muscle index of the limbs. The above conditions are difficult to meet for small institutions, community hospitals, and home environments that lack the necessary equipment and personnel. In addition, skeletal muscle measurement methods similar to DEXA have the problem of radiation exposure, and cannot achieve real-time, dynamic, low-cost, and self-assessment monitoring, which poses a serious challenge to the early detection and early intervention of sarcopenia. Because of this, there is an urgent need to develop a system that does not require physical measurement, is fast, self-assessed, and can be deployed online to promote early screening, early intervention, and dynamic monitoring of sarcopenia. The most ideal tool is one that does not require individuals to undergo any physical measurements that rely on equipment or professionals, and the evaluation results of sarcopenia can be obtained only through questions and answers or self-assessment. Summary of the invention

[0006] In view of this, the purpose of the present invention is to provide a method and system for constructing a sarcopenia prediction model without physical measurement, so as to serve as an auxiliary decision-making tool to realize the risk assessment of sarcopenia in various scenarios such as low cost, remote and self-assessment.

[0007] In order to achieve the above object, the present invention provides the following technical solutions:

[0008] A method for constructing a sarcopenia prediction model without physical measurement, comprising the following steps:

[0009] Step 1: Establish a population data set: Select community population or inpatient population, and collect variable information of the modeling population through public databases or medical electronic medical record systems, including but not limited to age, gender, height, weight, limb skeletal muscle index, dominant hand grip strength, five sit-to-stand tests, six-meter walking speed test, total score and sub-item results of daily living activities questionnaire, total score and sub-item results of instrumental daily living activities questionnaire as data set samples;

[0010] Step 2: Sample data preprocessing: Preprocess the collected sample data sets, including but not limited to deleting outliers, excluding samples with missing key modeling information, interpolating missing data using multiple interpolation, unique-hot encoding of multi-classified data, standardizing continuous variables, and deleting variables with the same values ​​for all samples; define the outcome label sarcopenia, and retrospectively diagnose sarcopenia using widely recognized international guidelines, including but not limited to the 2019 guidelines of the Asian Sarcopenia Working Group and the guidelines of the European Elderly Sarcopenia Working Group; encode people with sarcopenia as 1, otherwise they are coded as 0;

[0011] Step 3: Divide the sample data set preprocessed in step 2 into a training set and a test set according to a certain ratio;

[0012] Step 4: Modeling feature screening: Select the training set divided in step 3, take sarcopenia as the outcome variable, and take the information that can be obtained only through question and answer after preprocessing in step 1 as the independent variable, including but not limited to age, gender, height, weight, whether it is difficult to complete a one-kilometer jog, and whether it is difficult to lift a 5-kilogram weight with the upper limbs; use a linear or nonlinear algorithm to perform feature screening on the independent variables in the training set, and achieve dimensionality reduction of the independent variables according to the characteristic coefficient, significance or importance ranking of the variables, retain the key features that have an important impact on the model prediction performance, and eliminate irrelevant features that have a small impact on the prediction results, and finally obtain a more concise and efficient feature subset;

[0013] Step 5: Construct multiple machine learning classifiers: Based on the training set after dimensionality reduction in step 4, a supervised machine learning algorithm is used to construct multiple classifiers. According to the number and value range of hyperparameters of different classifiers, the hyperparameters of the classifiers are optimized by random search or grid search. A variety of evaluation graphs and indicators are generated by cross-validation or validation in the validation set, including but not limited to receiver operating characteristic curves, calibration curves, areas under the curves and 95% confidence intervals, and Brier scores.

[0014] Step 6: Select the best model: Based on the model evaluation index obtained in step 5, select a main evaluation index according to the purpose, sort the candidate classifiers according to the index evaluation index, and select the classifier with the best performance as the best model;

[0015] Step 7: Performance evaluation and prediction: Based on the best model obtained in step 6, the occurrence of sarcopenia in the test set samples or samples manually input by the user is quickly predicted.

[0016] Furthermore, in step one, a specific subgroup of the population is selected for stratified modeling to reduce the amount of calculation and obtain a more targeted model.

[0017] Furthermore, in step 2, the standardization method of continuous variables is the z-score method.

[0018] Furthermore, in step 4, LASSO (Least Absolute Shrinkage and Selection Operator) method is used for feature screening.

[0019] Furthermore, in step five, a variety of classifiers are constructed using supervised machine learning algorithms, including but not limited to Logistic regression model, Gaussian naive Bayes model, gradient boosting machine model, random forest algorithm, multi-layer perceptron, extreme gradient boosting machine, lightweight gradient boosting machine model and K nearest neighbor algorithm.

[0020] Furthermore, in step six, the Brier score or the area under the curve of the test set is used as an evaluation indicator.

[0021] A sarcopenia prediction system based on the method without physical measurement comprises:

[0022] Data collection module: including a model training data input submodule and an individual prediction data input submodule, which are respectively used to collect the training data required for modeling from the electronic medical data system, local data source or other data source, and receive single or batch model independent variable information input by the user as input data;

[0023] Data analysis module: including a prediction model building submodule and a prediction execution submodule, which are respectively used to use the data obtained by the model training data input submodule to execute the modeling steps of step 1 to step 7, obtain a sarcopenia prediction model and model performance evaluation indicators without physical measurement, and the sarcopenia prediction model obtained by the prediction model building submodule, use the input data obtained by the data acquisition module to perform risk calculation, and obtain the prediction probability and label category of each category of sarcopenia;

[0024] Result output module: includes a batch output submodule and a case output submodule, which are respectively used to output model performance evaluation indicators according to the modeling results of the prediction model modeling submodule, and to output the predicted probability and label category of each category of sarcopenia according to the calculation results of the execution prediction submodule.

[0025] The beneficial effects of the present invention are:

[0026] (1) Compared with the commonly used sarcopenia judgment method, the present invention does not use any physical measurement information that needs to be obtained through instruments, equipment or guidance of health professionals in the prediction process. All independent variable information can be obtained only through questions and answers or individual self-assessment to complete the risk assessment of sarcopenia. Therefore, it is expected to be widely applicable to various types of hospitals at all levels, community health service institutions, family scenarios and research scenarios, and has the characteristics of rapidity and low cost;

[0027] (2) The present invention does not specify a specific machine learning classification algorithm, so a specific algorithm can be reasonably selected according to different populations, disease stages, data types, and sample sizes. This improves the flexibility of the present invention in the application process and can meet the needs of saving computing costs under different hardware conditions;

[0028] (3) The present invention can achieve graded prediction of the severity of sarcopenia by replacing the multi-classification machine learning algorithm in the modeling step;

[0029] (4) The present invention is expected to be applicable to other systems that use muscle mass loss as the main diagnostic criterion and has a certain degree of derivability.

[0030] Other advantages, objectives and features of the present invention will be described in the following description to some extent, and to some extent, will be obvious to those skilled in the art based on the following examination and study, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below in conjunction with the accompanying drawings, wherein:

[0032] Figure 1 It is a technical flow chart of the present invention;

[0033] Figure 2 This is an analysis diagram for feature screening in Example 1 of the present invention;

[0034] Figure 3 is a calibration curve diagram of different models in Example 1 of the present invention;

[0035] Figure 4 is a receiver operating characteristic curve diagram of the best model in Example 1 of the present invention;

[0036] Figure 5 This is an analysis diagram for feature screening in Example 2 of the present invention;

[0037] Figure 6Receiver operating characteristic curves of different models in Example 2 of the present invention;

[0038] Figure 7 This is a calibration curve diagram of the best model in Example 2 of the present invention. DETAILED DESCRIPTION

[0039] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner, and the following embodiments and the features in the embodiments can be combined with each other without conflict.

[0040] Among them, the drawings are only used for illustrative explanations, and they only represent schematic diagrams rather than actual pictures, and should not be understood as limitations on the present invention. In order to better illustrate the embodiments of the present invention, some parts of the drawings may be omitted, enlarged or reduced, and do not represent the size of actual products. For those skilled in the art, it is understandable that some well-known structures and their descriptions in the drawings may be omitted.

[0041] The same or similar numbers in the drawings of the embodiments of the present invention correspond to the same or similar parts; in the description of the present invention, it should be understood that if the terms "upper", "lower", "left", "right", "front", "rear", etc. indicate the orientation or position relationship, they are based on the orientation or position relationship shown in the drawings, which is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation. Therefore, the terms describing the position relationship in the drawings are only used for illustrative purposes and cannot be understood as limiting the present invention. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.

[0042] In a first aspect, the present invention provides a method for constructing a sarcopenia prediction model without physical measurement, comprising the following steps:

[0043] 1) Establish a crowd data set:

[0044] Select community population or inpatient population, and collect variable information of modeling population through public databases or medical electronic medical record systems, including but not limited to age, gender, height, weight, limb skeletal muscle index, dominant hand grip strength, five sit-to-stand tests, six-meter walking speed test, total score and sub-item results of daily living activities questionnaire, total score and sub-item results of instrumental daily living activities questionnaire as data set samples;

[0045] 2) Sample data preprocessing:

[0046] Preprocess the collected sample data sets, including but not limited to deleting outliers, excluding samples with missing key modeling information, interpolating missing data using multiple interpolation, unique-hot encoding of multi-classified data, standardizing continuous variables, and deleting variables with the same values ​​for all samples; define the outcome label sarcopenia, and retrospectively diagnose sarcopenia using widely recognized international guidelines, including but not limited to the 2019 guidelines of the Asian Sarcopenia Working Group and the guidelines of the European Elderly Sarcopenia Working Group; encode people with sarcopenia as 1, otherwise as 0. If severity grading is to be performed, encode people without sarcopenia as 0, sarcopenia as 1, and severe sarcopenia as 2;

[0047] 3) Divide the training set and test set:

[0048] The sample data set preprocessed in step 2) is randomly divided into a training set and a test set according to a certain ratio. The training set occupies a larger proportion and is used for model training, while the training set data occupies a smaller proportion and is used to evaluate the effectiveness of model classification;

[0049] 4) Modeling feature screening

[0050] The training set divided in step 3) is selected, and sarcopenia is used as the outcome variable. The information that can be obtained only through question and answer after preprocessing in step 1) is used as the independent variable, including but not limited to age, gender, height, weight, whether it is difficult to complete a one-kilometer jog, and whether it is difficult to lift a 5-kilogram weight with the upper limbs; a linear or nonlinear algorithm is used to perform feature screening on the independent variables in the training set, and the dimension reduction of the independent variables is achieved according to the characteristic coefficient, significance or importance ranking of the variables, retaining the key features that have an important impact on the prediction performance of the model, and eliminating irrelevant features that have a small impact on the prediction results, so as to finally obtain a more concise and efficient feature subset;

[0051] 5) Build multiple machine learning classifiers:

[0052] Based on the training set after dimension reduction in step 4), a plurality of classifiers are constructed by using a supervised machine learning algorithm, and the hyperparameters of the classifiers are optimized by using a random search or a grid search method according to the number and value range of the hyperparameters of different classifiers, and a plurality of evaluation graphs and indicators are generated by using a cross-validation method or a validation method in a validation set, including but not limited to a receiver operating characteristic curve, a calibration curve, an area under the curve and a 95% confidence interval, and a Brier score;

[0053] 6) Select the best model

[0054] Based on the model evaluation index obtained in step 5), a main evaluation index is selected according to the purpose, the candidate classifiers are sorted according to the index evaluation index, and the classifier with the best performance is selected as the best model;

[0055] 7) Performance evaluation and prediction

[0056] According to the best model obtained in step 6), the occurrence of sarcopenia in the test set samples or samples manually input by the user is quickly predicted;

[0057] In step 1), a specific subgroup of the population is selected for stratified modeling to reduce the amount of calculation and obtain a more targeted model;

[0058] In step 2), the standardization method of continuous variables is the z-score method, and the specific formula is: (x-μ) / σ. Wherein μ represents the mean of the overall data, σ represents the standard deviation of the overall data, and x represents the individual observation value;

[0059] In the step 4), LASSO (Least Absolute Shrinkage and Selection Operator) method is used for feature screening.

[0060] In a second aspect, the present invention provides a sarcopenia prediction system without physical measurement, see Figure 1 The system mainly includes the following modules:

[0061] Data acquisition module:

[0062] 1) Model training data input submodule: used to collect training data required for modeling from electronic medical data systems, local data sources or other data sources, including but not limited to age, gender, height, weight, limb skeletal muscle index, dominant hand grip strength, five sit-to-stand tests, six-meter walking speed test, daily living activities questionnaire total score and sub-item results, instrumental daily living activities questionnaire total score and sub-item results, disease information, treatment information, nutrition-related anthropometric indicators, hematological indicators, short-term and long-term clinical outcomes as input data;

[0063] 2) Individual prediction data input submodule: receiving single or batch model independent variable information input by the user as input data;

[0064] Data Analysis Module:

[0065] 1) Prediction model building submodule: using the data obtained by the model training data input submodule, executing the modeling steps of step 1) to step 7) in the first aspect, and obtaining a sarcopenia prediction model and model performance evaluation index without physical measurement;

[0066] 2) executing the prediction submodule: performing risk calculation based on the sarcopenia prediction model obtained by the prediction model building submodule, using the input data obtained by the data acquisition module, and obtaining the prediction probability and label category of each category of sarcopenia;

[0067] Result output module:

[0068] 1) Batch output submodule: output model performance evaluation indicators according to the modeling results of the prediction model modeling submodule;

[0069] 2) Case output submodule: outputs the predicted probability and label category of each category of sarcopenia according to the calculation results of the execution prediction submodule.

[0070] Embodiment 1:

[0071] In this embodiment, see Figure 1 The data analysis module and the result output module in the method for constructing a sarcopenia prediction model without physical measurement include the following steps:

[0072] 1) Establish a crowd data set:

[0073] 10,246 cases of community-dwelling population data were selected, and individual data including age, gender, height, weight, whether there is difficulty in dressing, whether there is difficulty in jogging 1 km, whether there is difficulty in walking 1 km, whether there is difficulty in lifting 5 kg, grip strength, limb skeletal muscle index, and five sit-stand tests were collected as data set samples;

[0074] 2) Sample data preprocessing:

[0075] The collected sample data sets were preprocessed, and the continuous variables were z-score standardized. The specific formula was: (x-μ) / σ. Where μ represents the mean of the overall data, σ represents the standard deviation of the overall data, and x represents the individual observation value. Sarcopenia was retrospectively diagnosed as the outcome label based on the 2019 criteria of the Asian Sarcopenia Working Group based on grip strength, limb skeletal muscle index, and five sit-to-stand tests. People with sarcopenia were coded as 1, otherwise coded as 0;

[0076] 3) Divide the training set and test set:

[0077] The sample data set preprocessed in step 2) is randomly divided into a training set and a test set in a ratio of 7:3;

[0078] 4) Modeling feature screening

[0079] Select the training set divided in step 3), take sarcopenia as the outcome variable, and use the preprocessed features in step 1) as independent variables to construct the LASSO method to rank the feature importance, retain all features with coefficient absolute values ​​greater than 0, and obtain a subset containing 8 features, see Figure 2 .

[0080] 5) Build multiple classifiers:

[0081] Based on the training set after dimension reduction in step 4), four classifiers were constructed using supervised machine learning algorithms in Python 3.9.11 environment, including Logistic regression model, Gaussian naive Bayes model, gradient boosting model and random forest algorithm. Random search was iterated 1000 times to find the hyperparameters of each model, generate calibration curves and calculate Brier scores, see Figure 3 .

[0082] 6) Select the best model

[0083] Based on the model evaluation index obtained in step 5), the Brier score of the test set is used as the main evaluation index, the candidate classifiers are sorted according to the index evaluation index, and the Gaussian naive Bayes model with the best performance is selected as the best model, see Figure 3 .

[0084] 7) Performance evaluation and prediction

[0085] According to the best model obtained in step 6), the occurrence of sarcopenia in the test set samples was quickly predicted, and the area under the curve and 95% confidence interval, accuracy, sensitivity, specificity, F1 score, positive predictive value and negative predictive value were calculated. Figure 4 .

[0086] Embodiment 2:

[0087] This embodiment is substantially the same as the first embodiment. Figure 1 The data analysis module and result output module in the above examples are different, but different feature screening strategies, different machine learning algorithm modeling, and different model evaluation indicators are used. A method for constructing a sarcopenia prediction model without physical measurement includes the following steps:

[0088] 1) Establish a crowd data set

[0089] 10,246 cases of community-dwelling population data were selected, and individual data including age, weight, difficulty in dressing, difficulty in jogging 1 km, difficulty in walking 1 km, difficulty in lifting 5 kg, grip strength, limb skeletal muscle index, and five sit-stand tests were collected as data set samples;

[0090] 2) Sample data preprocessing

[0091] The collected sample data sets were preprocessed, and the continuous variables were z-score standardized. The specific formula was: (x-μ) / σ. Where μ represents the mean of the overall data, σ represents the standard deviation of the overall data, and x represents the individual observation value. Sarcopenia was retrospectively diagnosed as the outcome label based on the 2019 criteria of the Asian Sarcopenia Working Group based on grip strength, limb skeletal muscle index, and five sit-to-stand tests. People with sarcopenia were coded as 1, otherwise coded as 0;

[0092] 3) Divide the training set and test set

[0093] The sample data set preprocessed in step 2) is randomly divided into a training set and a test set in a ratio of 7:3;

[0094] 4) Modeling feature screening

[0095] Select the training set divided in step 3), take sarcopenia as the outcome variable, and use the preprocessed features in step 1) as independent variables to construct the LASSO method to rank the feature importance, select the top 4 features with the absolute value of the feature coefficient, and obtain a subset containing 4 features, see Figure 5 .

[0096] 5) Build multiple classifiers

[0097] Based on the training set after dimension reduction in step 4), four classifiers were constructed using supervised machine learning algorithms in Python 3.9.11 environment, including multilayer perceptron, extreme gradient boosting machine, lightweight gradient boosting machine model and K nearest neighbor algorithm. The hyperparameters of each model were found by random search iteration 1000 times, and the receiver operating characteristic curve was generated and the area under the curve and 95% confidence interval were calculated. See Figure 6 .

[0098] 6) Select the best model

[0099] Based on the model evaluation index obtained in step 5), the area under the curve of the test set is selected as the main evaluation index, the candidate classifiers are sorted according to the index evaluation index, and the multilayer perceptron model with the best performance is selected as the best model. Figure 6 .

[0100] 7) Performance evaluation and prediction

[0101] According to the best model obtained in step 6), the occurrence of sarcopenia in the test set samples was quickly predicted, a calibration curve was drawn, and the Brier score, accuracy, sensitivity, specificity, F1 score, positive predictive value and negative predictive value were calculated. Figure 7 .

[0102] From the results, it can be seen that the present invention uses the variable information obtained through question-and-answer or self-assessment to achieve high-performance prediction of the risk of sarcopenia without any physical measurement. In Example 1 and Example 2, the present invention constructs a sarcopenia prediction model without physical measurement, which is expected to be used as an auxiliary decision-making tool to achieve low-cost, remote, self-assessment and other scenarios for sarcopenia risk assessment.

[0103] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution, which should be included in the scope of the claims of the present invention.

Claims

1. A method for constructing a sarcopenia prediction model without physical measurement, characterized in that: The following steps are involved: Step 1: Establish a population data set: Select community population or inpatient population, and collect variable information of the modeling population through public databases or medical electronic medical record systems, including but not limited to age, gender, height, weight, limb skeletal muscle index, dominant hand grip strength, five sit-to-stand tests, six-meter walking speed test, total score and sub-item results of daily living activities questionnaire, total score and sub-item results of instrumental daily living activities questionnaire as data set samples; Step 2: Sample data preprocessing: Preprocess the collected sample data sets, including but not limited to deleting outliers, excluding samples with missing key modeling information, interpolating missing data using multiple interpolation, unique-hot encoding of multi-classified data, standardizing continuous variables, and deleting variables with the same values ​​for all samples; define the outcome label sarcopenia, and retrospectively diagnose sarcopenia using widely recognized international guidelines, including but not limited to the 2019 guidelines of the Asian Sarcopenia Working Group and the guidelines of the European Elderly Sarcopenia Working Group; encode people with sarcopenia as 1, otherwise they are coded as 0; Step 3: Divide the sample data set preprocessed in step 2 into a training set and a test set according to a certain ratio; Step 4: Modeling feature screening: Select the training set divided in step 3, take sarcopenia as the outcome variable, and take the information that can be obtained only through question and answer after preprocessing in step 1 as the independent variable, including but not limited to age, gender, height, weight, whether it is difficult to complete a one-kilometer jog, and whether it is difficult to lift a 5-kilogram weight with the upper limbs; use a linear or nonlinear algorithm to perform feature screening on the independent variables in the training set, and achieve dimensionality reduction of the independent variables according to the characteristic coefficient, significance or importance ranking of the variables, retain the key features that have an important impact on the model prediction performance, and eliminate irrelevant features that have a small impact on the prediction results, and finally obtain a more concise and efficient feature subset; Step 5: Construct multiple machine learning classifiers: Based on the training set after dimensionality reduction in step 4, a supervised machine learning algorithm is used to construct multiple classifiers. According to the number and value range of hyperparameters of different classifiers, the hyperparameters of the classifiers are optimized by random search or grid search. A variety of evaluation graphs and indicators are generated by cross-validation or validation in the validation set, including but not limited to receiver operating characteristic curves, calibration curves, areas under the curves and 95% confidence intervals, and Brier scores. Step 6: Select the best model: Based on the model evaluation index obtained in step 5, select a main evaluation index according to the purpose, sort the candidate classifiers according to the index evaluation index, and select the classifier with the best performance as the best model; Step 7: Performance evaluation and prediction: Based on the best model obtained in step 6, the occurrence of sarcopenia in the test set samples or samples manually input by the user is quickly predicted.

2. The method for constructing a sarcopenia prediction model without physical measurement according to claim 1, characterized in that: In the step 1, a specific subgroup of the population is selected for stratified modeling to reduce the amount of calculation and obtain a more targeted model.

3. The method for constructing a sarcopenia prediction model without physical measurement according to claim 1, characterized in that: In the step 2, the standardization method of continuous variables is the z-score method.

4. The method for constructing a sarcopenia prediction model without physical measurement according to claim 1, characterized in that: In the step 4, LASSO (Least Absolute Shrinkage and Selection Operator) method is used for feature screening.

5. The method for constructing a sarcopenia prediction model without physical measurement according to claim 1, characterized in that: In the step 5, a variety of classifiers are constructed using supervised machine learning algorithms, including but not limited to Logistic regression model, Gaussian naive Bayes model, gradient boosting machine model, random forest algorithm, multi-layer perceptron, extreme gradient boosting machine, lightweight gradient boosting machine model and K nearest neighbor algorithm.

6. The method for constructing a sarcopenia prediction model without physical measurement according to claim 1, characterized in that: In step six, the Brier score or area under the curve of the test set is used as an evaluation indicator.

7. A sarcopenia prediction system without physical measurement based on the method according to any one of claims 1 to 6, characterized in that: include: Data collection module: including a model training data input submodule and an individual prediction data input submodule, which are respectively used to collect the training data required for modeling from the electronic medical data system, local data source or other data sources, and receive single or batch model independent variable information input by the user as input data; Data analysis module: including a prediction model building submodule and a prediction execution submodule, which are respectively used to use the data obtained by the model training data input submodule to execute the modeling steps of step 1 to step 7, obtain a sarcopenia prediction model and model performance evaluation indicators without physical measurement, and the sarcopenia prediction model obtained by the prediction model building submodule, use the input data obtained by the data acquisition module to perform risk calculation, and obtain the prediction probability and label category of each category of sarcopenia; Result output module: includes a batch output submodule and a case output submodule, which are respectively used to output model performance evaluation indicators according to the modeling results of the prediction model modeling submodule, and to output the predicted probability and label category of each category of sarcopenia according to the calculation results of the execution prediction submodule.

Images