Bone mass loss prediction system, model establishment system and computer equipment

Abstract

The invention relates to a bone mass loss prediction system, a model establishment system, computer equipment, a storage medium and a computer program product. The system comprises a sample data obtaining and grouping module which obtains a preset number of bone mass data and influence factor data sets corresponding to the bone mass data, knows a bone mass normal group, a bone mass reduction group and an osteoporosis group which are divided by the bone mass data according to a DXA examination result, and obtains a reduction sample set and an osteoporosis sample set; the model training module is used for training to obtain a bone mass reduction risk prediction model and an osteoporosis risk prediction model; and the evaluation and verification module generates a bone mass loss visual display tool and evaluates and verifies the bone mass reduction risk prediction model and the osteoporosis risk prediction model to obtain a bone mass loss prediction model. By adopting the system, whether a T2DM patient has bone mass loss or not can be quickly identified.

Description

Technical Field

[0001] The present application relates to the technical field of type 2 diabetes complication prediction, and in particular to a bone loss prediction system, a model building system, a computer device, a storage medium and a computer program product. Background Art

[0002] Type 2 diabetes mellitus (T2DM) is a chronic metabolic disease characterized by elevated blood sugar, and the number of patients has increased dramatically worldwide. Studies have found that there is a complex relationship between diabetes and osteoporosis. Type 2 diabetes is considered an independent risk factor for osteoporosis and fragility fractures. The skeletal system complications of T2DM patients are an important part of the comprehensive management of diabetes. It is urgent to strengthen the early screening and prevention of bone loss in T2DM patients.

[0003] Traditional prediction models only predict the risk of osteoporosis, but do not predict people with osteopenia who are also at risk of fracture. Therefore, a simple and convenient tool that can be widely used in clinical practice is needed to predict the risk of bone loss in patients with T2DM. Summary of the invention

[0004] Based on this, it is necessary to provide a bone loss prediction system, model building system, computer equipment, storage medium and computer program product that can quickly identify bone loss in T2DM patients at an early stage to address the above technical problems.

[0005] In a first aspect, the present application provides a bone loss prediction system, the system comprising:

[0006] A reference data acquisition module is used to obtain a reference influencing factor data set of the object to be tested;

[0007] The result data output module is used to call the pre-established bone loss prediction model, input the reference influencing factor data set of the object to be tested, and obtain the bone loss prediction result of the object to be tested.

[0008] In a second aspect, the present application provides a system for establishing a bone loss prediction model, for establishing a bone loss prediction model as provided in the first aspect, the system comprising:

[0009] The sample data acquisition and grouping module is used to obtain a preset number of bone mass data and the influencing factor data sets corresponding to each bone mass data, and to know the normal bone mass group, osteopenia group and osteoporosis group into which the bone mass data are divided according to the DXA (Dual-emission X-ray Absorptiometry) examination results, to obtain each osteopenia probability and its corresponding influencing factor data set, and to form a single osteopenia sample with a single osteopenia probability and its corresponding influencing factor data set to obtain a osteopenia sample set; to obtain each osteoporosis probability and its corresponding influencing factor data set, and to form a single osteoporosis sample with a single osteoporosis probability and its corresponding influencing factor data set to obtain a osteoporosis sample set;

[0010] A model training module is used to train a risk prediction model for osteopenia by taking a dataset of influencing factors in a reduction sample as input and a probability of osteopenia in a reduction sample as output; and to train a risk prediction model for osteoporosis by taking a dataset of influencing factors in a loose sample as input and a probability of osteoporosis in a loose sample as output;

[0011] The evaluation and verification module is used to generate a bone loss visualization display tool based on the osteopenia risk prediction model and the osteoporosis risk prediction model, and to evaluate and verify the osteopenia risk prediction model and the osteoporosis risk prediction model to obtain a bone loss prediction model with a reference influencing factor data set as input and a bone loss prediction result as output; wherein the bone loss prediction model includes the osteopenia risk prediction model and the osteoporosis risk prediction model after evaluation and verification.

[0012] In a third aspect, the present application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, it implements the functions of the above-mentioned bone loss prediction system; or, implements the functions of the above-mentioned bone loss prediction model establishment system.

[0013] In a fourth aspect, the present application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the functions of the above-mentioned bone loss prediction system; or, implements the functions of the above-mentioned bone loss prediction model establishment system.

[0014] In a fifth aspect, the present application also provides a computer program product, including a computer program, which, when executed by a processor, implements the functions of the above-mentioned bone loss prediction system; or, implements the functions of the above-mentioned bone loss prediction model establishment system.

[0015] The above-mentioned bone loss prediction system, model building system, computer equipment, storage medium and computer program product, the reference data acquisition module obtains the reference influencing factor data set of the object to be tested, the result data output module calls the pre-established bone loss prediction model, inputs the reference influencing factor data set of the object to be tested, and obtains the bone loss prediction result of the object to be tested, so as to achieve early detection and early intervention of people with T2DM patients complicated with osteopenia, thereby reducing the risk of osteoporosis and even fractures, and improving the survival and quality of life of patients. The sample data acquisition and grouping module obtains a preset number of bone mass data and the influencing factor data sets corresponding to each bone mass data, and knows the bone mass normal group, osteopenia group and osteoporosis group divided by the DXA examination results, obtains each bone mass reduction probability and its corresponding influencing factor data set, and forms a single reduction sample with a single bone mass reduction probability and its corresponding influencing factor data set, and obtains a reduction sample set; obtains each osteoporosis probability and its corresponding influencing factor data set, and forms a single loose sample with a single osteoporosis probability and its corresponding influencing factor data set, and obtains a loose sample set; the model The training module takes the influencing factor dataset in the reduction sample as input and the probability of osteopenia in the reduction sample as output, and obtains the osteopenia risk prediction model through training; takes the influencing factor dataset in the loose sample as input and the probability of osteoporosis in the loose sample as output, and obtains the osteoporosis risk prediction model through training; the evaluation and verification module generates a bone loss visualization display tool based on the osteopenia risk prediction model and the osteoporosis risk prediction model, and evaluates and verifies the osteopenia risk prediction model and the osteoporosis risk prediction model, and obtains a bone loss prediction model with the reference influencing factor dataset as input and the bone loss prediction result as output. This model provides clinical medical staff with an evaluation tool that can quickly identify whether T2DM patients have bone loss in the early stage, and has the advantages of being portable and economical. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A schematic diagram of modules of a bone loss prediction system in one embodiment;

[0017] Figure 2 A schematic diagram of modules of a system for establishing a bone loss prediction model in one embodiment;

[0018] FIG3( a ) is a LASSO regression variable shrinkage coefficient diagram of an osteopenia risk prediction model in one embodiment;

[0019] FIG3( b ) is a graph showing the relationship between the LASSO regression ten-fold cross validation model error and Log(λ) of the bone mass reduction risk prediction model in one embodiment;

[0020] FIG4( a ) is a LASSO regression variable shrinkage coefficient diagram of an osteoporosis risk prediction model in one embodiment;

[0021] FIG4( b ) is a graph showing the relationship between the LASSO regression ten-fold cross validation model error and Log(λ) of an osteoporosis risk prediction model in one embodiment;

[0022] Figure 5 A schematic diagram of a nomogram model for T2DM patients with concurrent osteopenia in one embodiment;

[0023] Figure 6 A schematic diagram of a nomogram model for T2DM patients with osteoporosis in one embodiment;

[0024] Figure 7 A schematic diagram of a web calculator for predicting bone mass loss risk in one embodiment;

[0025] Figure 8 A schematic diagram of an osteoporosis risk prediction model web calculator in one embodiment;

[0026] FIG9( a ) is a graph showing a receiver operating characteristic curve of an osteopenia risk prediction model in one embodiment;

[0027] FIG9( b ) is a graph showing a receiver operating characteristic curve of an osteoporosis risk prediction model in one embodiment;

[0028] FIG10( a ) is an internal validation diagram of a receiver operating characteristic curve of an osteopenia risk prediction model in one embodiment;

[0029] FIG10( b ) is an internal validation diagram of a receiver operating characteristic curve of an osteoporosis risk prediction model in one embodiment;

[0030] FIG11( a ) is a comparison diagram of a nomogram of a risk prediction model for osteopenia and a receiver operating characteristic curve of a single risk factor in one embodiment;

[0031] FIG11( b ) is a comparison diagram of the osteoporosis risk prediction model nomogram and the receiver operating characteristic curve of a single risk factor in one embodiment;

[0032] FIG. 12( a ) is an internal validation diagram of a calibration line and calibration degree of a bone mass reduction risk prediction model in one embodiment;

[0033] FIG12( b ) is an internal validation diagram of the calibration line and calibration degree of the osteoporosis risk prediction model in one embodiment;

[0034] FIG13( a ) is a comparison diagram of a nomogram of an osteopenia risk prediction model and a clinical decision curve diagram of a single risk factor in one embodiment;

[0035] FIG13( b ) is a comparison diagram of a nomogram of an osteoporosis risk prediction model and a clinical decision curve diagram of a single risk factor in one embodiment;

[0036] Fig.14 A schematic diagram of a process for predicting bone loss in one embodiment;

[0037] Fig.15 A schematic diagram of a process for establishing a bone loss prediction model in one embodiment;

[0038] Fig.16 A schematic diagram of a process for establishing a bone loss prediction model in another embodiment;

[0039] Fig.17 FIG. 4 is a flow chart of a method for predicting bone loss in another embodiment. DETAILED DESCRIPTION

[0040] Type 2 diabetes is a chronic metabolic disease characterized by elevated blood sugar levels, which can lead to a variety of acute and chronic complications. The number of patients has increased dramatically worldwide. In recent years, diabetes-related bone complications have also received increasing attention. Studies have shown that there is a complex relationship between diabetes and osteoporosis, and type 2 diabetes has been considered an independent risk factor for osteoporosis and fragility fractures. Complications of the skeletal system in patients with T2DM are an important part of the comprehensive management of diabetes. It is urgent to strengthen early screening and prevention of concurrent bone loss in patients with T2DM.

[0041] Bone loss includes two stages: osteopenia and osteoporosis. Osteopenia (OS) is the early stage of osteoporosis. Studies have shown that 17.7% of patients with osteopenia will eventually develop osteoporosis. Osteoporosis (OP) is an abnormal bone metabolism disease characterized by a decrease in the content of bone tissue cells per unit volume, disordered bone metabolism, and destruction of bone structure, which leads to increased bone brittleness and susceptibility to pathological fractures. OP has become an important health problem for people over 50 years old in my country. Osteoporosis is particularly serious in postmenopausal women, and the OS population is also quite large. OP has been recognized by the World Health Organization as a chronic non-communicable disease that seriously endangers human health after cardiovascular disease. The hidden nature of the pathogenesis of osteoporosis poses a challenge to the early identification and early intervention of osteoporosis. Osteoporosis gradually evolves from osteopenia. The healthcare system needs to propose effective screening and intervention policies to intervene as early as possible and delay the progression of osteopenia, ultimately reducing the risk of osteoporosis and even fractures.

[0042] The application of traditional osteoporosis-related assessment scales in patients with type 2 diabetes needs further study. Traditional prediction models only predict the risk of osteoporosis, but not the large population of people with osteopenia who are also at risk of fractures. Therefore, it is necessary to develop a simple and convenient tool that can be widely used in clinical practice to predict the risk of bone loss in patients with T2DM.

[0043] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0044] In one embodiment, Figure 1 As shown, a bone loss prediction system is provided, the system comprising:

[0045] A reference data acquisition module 102 is used to acquire a reference influencing factor data set of the object to be tested;

[0046] Specifically, the reference influencing factor data set of the test subject includes age, gender, weight, systolic blood pressure, duration of diabetes, blood creatinine, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, β-CTX, 25(OH)VD, parathyroid hormone, drinking history, smoking history, ITP1NP, etc., among which β-CTX represents a special sequence of β-collagen, 25(OH)VD represents 25-hydroxyvitamin D, and TP1NP represents total type I procollagen amino-terminal extended peptide.

[0047] The result data output module 104 is used to call the pre-established bone loss prediction model, input the reference influencing factor data set of the subject to be tested, and obtain the bone loss prediction result of the subject to be tested.

[0048] Specifically, the reference influencing factor data set of the subject to be tested is input into the bone loss prediction model to obtain the prediction data of bone loss or osteoporosis in T2DM patients.

[0049] In the above-mentioned bone loss prediction system, the reference influencing factor data set of the object to be tested is obtained through the reference data acquisition module, and then the pre-established bone loss prediction model is called through the result data output module. The reference influencing factor data set of the object to be tested is input, and the bone loss prediction result of the object to be tested can be obtained. This is convenient and fast, which is conducive to clinical medical staff to predict and evaluate the probability of bone loss in T2DM patients, and realize early detection and early intervention of T2DM patients with concurrent bone loss.

[0050] In one embodiment, Figure 2As shown, a bone loss prediction model establishment system is provided, which is used to establish the bone loss prediction model provided in the above embodiment, including:

[0051] The sample data acquisition and grouping module 202 is used to obtain a preset number of bone mass data and the influencing factor data sets corresponding to each bone mass data, and to know the normal bone mass group, osteopenia group and osteoporosis group into which the bone mass data are divided according to the DXA examination results, to obtain each bone mass reduction probability and its corresponding influencing factor data set, and to form a single reduction sample with a single bone mass reduction probability and its corresponding influencing factor data set to obtain a reduction sample set; to obtain each osteoporosis probability and its corresponding influencing factor data set, and to form a single loose sample with a single osteoporosis probability and its corresponding influencing factor data set to obtain a loose sample set.

[0052] Among them, the bone mass data of T2DM patients hospitalized in the endocrinology department and the corresponding influencing factor data sets were collected retrospectively, including the general information, clinical data and DXA examination results of T2DM patients.

[0053] Specifically, the influencing factor data set includes multiple possible risk factors. Optionally, the possible risk factors include gender, age (unit: years), height (unit: cm), weight (unit: kg), BMI (Body Mass Index), waist circumference (unit: cm), systolic blood pressure, diastolic blood pressure, smoking history, drinking history, history of hypertension, diabetes duration (unit: year), fasting blood glucose, fasting insulin, HOMA-IR (insulin resistance index), glycosylated hemoglobin, cholinesterase, total bilirubin, albumin, prealbumin, serum creatinine, serum uric acid, cystatin C, triglycerides, total cholesterol, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, urine microalbumin, N-MID (N-midosteocalcin), β-CTX, TP1NP, 25 (OH) VD, parathyroid hormone, hemoglobin, NLR (neutrophil to lymphocyte ratio), PLR (platelet to lymphocyte ratio), etc.

[0054] Optionally, the influencing factor data set is divided into multiple influencing factor data units according to specified standards, and the proportion of patients with normal bone mass, osteopenia, and osteoporosis in each influencing factor data unit is calculated, that is, the probability of normal bone mass, the probability of osteopenia, and the probability of osteoporosis; the specified standard is not unique and depends on the actual situation. For example, the influencing factor data set is represented in the form of vectors, and the distance between the vectors is calculated. The influencing factors with a distance within a certain range are divided into influencing factor data units.

[0055] Exemplarily, the patients are grouped according to the DXA examination results, that is, the bone mass data are grouped into a normal bone mass group, an osteopenia group, and an osteoporosis group.

[0056] Optionally, retrospective data collection inevitably has some missing data. Missing variables are considered to be randomly missing. IBM SPSS27.0 software is used to analyze the data, observe and clarify the maximum and minimum values ​​of each data and the missing data, and observe whether there are abnormal maximum or minimum values.

[0057] The model training module 204 is used to train a risk prediction model for osteopenia by taking the influencing factor dataset in the reduction sample as input and the probability of osteopenia in the reduction sample as output; and to train a risk prediction model for osteoporosis by taking the influencing factor dataset in the loose sample as input and the probability of osteoporosis in the loose sample as output;

[0058] Specifically, a risk prediction model for T2DM complicated with osteopenia was constructed based on the subjects in the normal bone mass group and the osteopenia group.

[0059] Specifically, a risk prediction model for T2DM complicated with osteoporosis was constructed based on the subjects in the osteopenia group and the osteoporosis group.

[0060] For example, R4.1.2 software was used to take multiple possible risk factors as independent variables, and potential risk factors with non-zero coefficients, i.e., potential predictive factors, were screened out through LASSO regression. The potential predictive factors screened out by LASSO regression were used as independent variables and included in multivariate Logistic regression analysis. The Backward method was used, and P < 0.05 was considered statistically significant. Finally, independent risk predictive factors were screened out, and bone loss risk prediction models and osteoporosis risk prediction models were established respectively.

[0061] The evaluation and verification module 206 is used to generate a bone loss visualization display tool based on the osteopenia risk prediction model and the osteoporosis risk prediction model, and evaluate and verify the osteopenia risk prediction model and the osteoporosis risk prediction model to obtain a bone loss prediction model trained with a reference influencing factor data set as input and a bone loss prediction result as output; wherein the bone loss prediction model includes the osteopenia risk prediction model and the osteoporosis risk prediction model after evaluation and verification.

[0062] Among them, the constructed prediction model is visualized using a nomogram and a web calculator.

[0063] Exemplarily, the osteopenia risk prediction model and the osteoporosis risk prediction model are respectively subjected to model discrimination evaluation and internal validation, model calibration evaluation and internal validation, and model clinical benefit evaluation.

[0064] Furthermore, the prediction results of the osteopenia risk prediction model and the data set of influencing factors of the osteoporosis risk prediction model can be used together as input data to train the osteoporosis risk prediction model, and the prediction results of the osteoporosis risk prediction model can be corrected; similarly, the prediction results of the osteoporosis risk prediction model and the data set of influencing factors of the osteopenia risk prediction model can be used together as input data to train the osteopenia risk prediction model, and the prediction results of the osteoporosis risk prediction model can be corrected.

[0065] Furthermore, the reference influencing factor data set of the subject to be tested is input into the bone loss prediction model to generate a bone reduction prediction result of the subject to be tested, and the bone reduction prediction result of the subject to be tested is compared with the target threshold; when the bone reduction prediction result of the subject to be tested is not higher than the target threshold, the bone loss prediction model outputs the bone reduction prediction result of the subject to be tested; when the bone reduction prediction result of the subject to be tested is higher than the target threshold, the bone loss prediction model will further output the osteoporosis prediction result of the subject to be tested.

[0066] In the above-mentioned bone loss prediction model establishment system, a preset number of bone mass data and the influencing factor data sets corresponding to each bone mass data are obtained through the sample data acquisition and grouping module, and the bone mass data are divided into normal bone mass group, osteopenia group and osteoporosis group according to the DXA examination results, and the probability of each bone loss and its corresponding influencing factor data set, each osteoporosis probability and its corresponding influencing factor data set are obtained, which can be modeled for the stages of osteopenia and osteoporosis respectively, and risk prediction can be carried out more targetedly. The osteopenia risk prediction model and the osteoporosis risk prediction model are obtained through training in the model training module, and the methods used are LASSO regression and Logistic regression analysis to achieve the optimal screening of variables. A bone loss visualization display tool is generated according to the osteopenia risk prediction model and the osteoporosis risk prediction model, and the osteopenia risk prediction model and the osteoporosis risk prediction model are evaluated and verified to obtain a bone loss prediction model with a reference influencing factor data set as input and a bone loss prediction result as output. The nomogram makes the results of the prediction model more readable, and displaying the prediction model through a web calculator is more conducive to the dissemination and use of the risk prediction model.

[0067] In an exemplary embodiment, the training of a prediction model for the risk of osteopenia includes: taking a data set of influencing factors in a sample as input and reducing the probability of osteopenia in the sample as output;

[0068] The influencing factor dataset in the reduced sample was used as the independent variable, and LASSO regression was used to screen the influencing factor dataset in the reduced sample according to the normal bone mass group and the osteopenia group to obtain the potential predictive factor set for osteopenia;

[0069] The potential predictive factor set of osteopenia was included in the logistic regression analysis to obtain the osteopenia predictive factor set and generate the osteopenia risk prediction model.

[0070] Exemplarily, the data set of influencing factors corresponding to the probability of bone loss in the reduced sample is used as the independent variable, and whether bone loss occurs is used as the dependent variable, where whether bone loss occurs can be the probability of bone loss in the reduced sample. Since there may be multicollinearity among various variables, 10-fold cross-validation LASSO regression is used to preliminarily screen the research variables in the data set of influencing factors corresponding to the probability of bone loss in the reduced sample to obtain a set of potential predictive factors for bone loss.

[0071] Specifically, the variables screened by LASSO regression were included in the multivariate logistic regression analysis to determine the predictors of osteopenia, namely, the osteopenia predictor set.

[0072] In this embodiment, the predictive factors of osteopenia are determined by LASSO regression analysis and multi-factor logistic regression analysis, thereby obtaining an osteopenia risk prediction model. LASSO regression is an algorithm of logistic regression, which introduces a penalty factor into the regression equation, can effectively deal with the multicollinearity problem between variables, avoid overfitting of data, and achieve optimal screening of variables.

[0073] In an exemplary embodiment, taking the influencing factor data set in the osteoporosis sample as input and the probability of osteoporosis in the osteoporosis sample as output, training to obtain an osteoporosis risk prediction model includes:

[0074] The influencing factor dataset in osteoporosis samples was used as the independent variable, and LASSO regression was used to screen the influencing factor dataset in osteoporosis samples according to the osteopenia group and osteoporosis group to obtain the potential predictive factor set of osteoporosis.

[0075] The osteoporosis potential predictive factor set was included in the Logistic regression analysis to obtain the osteoporosis predictive factor set and generate the osteoporosis risk prediction model.

[0076] Exemplarily, the influencing factor data set in the porous sample is used as the independent variable, and whether osteoporosis occurs is used as the dependent variable, wherein whether osteoporosis occurs can be the probability of osteoporosis in the porous sample. The LASSO regression with 10-fold cross-validation is used to perform preliminary screening of the research variables in the influencing factor data set to obtain a set of potential predictive factors for osteoporosis.

[0077] Specifically, the variables screened by LASSO regression were included in the multivariate logistic regression analysis to determine the predictors of osteoporosis, namely the osteoporosis predictor set.

[0078] In this embodiment, the prediction factors of osteoporosis are determined by LASSO regression analysis and multi-factor Logistic regression analysis, so as to obtain an osteoporosis risk prediction model and realize the risk prediction of osteoporosis.

[0079] In an exemplary embodiment, generating a bone loss visualization display tool according to an osteopenia risk prediction model and an osteoporosis risk prediction model includes:

[0080] The target software generates an osteopenia risk prediction nomogram and web calculator, an osteoporosis risk prediction nomogram and web calculator based on the osteopenia risk prediction model and the osteoporosis risk prediction model.

[0081] Exemplarily, the "rms" package in R 4.1.2 software is used, and the independent risk prediction factors finally screened out, namely, the osteopenia risk prediction model and the osteoporosis risk prediction model, are used to establish nomograms for predicting the risk of T2DM complicated by osteopenia and T2DM complicated by osteoporosis, namely, the osteopenia risk prediction nomogram and the osteoporosis risk prediction nomogram; the "DynNom" package is used to establish an osteopenia risk prediction web calculator and an osteoporosis risk prediction web calculator.

[0082] In this embodiment, the osteopenia risk prediction model and the osteoporosis risk prediction model are used in the form of a nomogram and a web calculator. Since the nomogram, also known as the Nomogram, is a complex regression equation converted into a visual graph, the results of the prediction model are more readable; at the same time, the prediction model can also be displayed in the form of a web calculator. By embedding the complete model formula into the background web page, the user only needs to enter the value of the patient's prediction variable to quickly calculate the accurate risk prediction value, which is more conducive to the dissemination and use of the risk prediction model. The web calculator can reduce the difficulty of using the prediction model in clinical practice, which is simpler and faster than using the nomogram model, and is easy to promote clinically.

[0083] In an exemplary embodiment, evaluating and validating the osteopenia risk prediction model and the osteoporosis risk prediction model includes:

[0084] The discrimination, calibration and clinical benefit of the osteopenia risk prediction model and osteoporosis risk prediction model were evaluated by specified methods and verified by the Bootstrap method.

[0085] In this embodiment, the osteopenia risk prediction model and the osteoporosis risk prediction model are evaluated and verified in different ways to ensure the discrimination, calibration and clinical benefit of the models, thereby ensuring the accuracy of the prediction results of the osteopenia risk prediction model and the osteoporosis risk prediction model.

[0086] In an exemplary embodiment, performing discrimination, calibration, and clinical benefit evaluation on the osteopenia risk prediction model and the osteoporosis risk prediction model by a specified method includes:

[0087] The receiver operating characteristic curve was used to evaluate the discrimination between the osteopenia risk prediction model and the osteoporosis risk prediction model.

[0088] The Homers-Lemeshow test was used to evaluate the calibration of the osteopenia risk prediction model and the osteoporosis risk prediction model;

[0089] The clinical decision curve was used to evaluate the clinical benefits of the osteopenia risk prediction model and osteoporosis risk prediction model.

[0090] Among them, the discrimination of the model refers to the ability of the prediction model to distinguish whether or not a certain disease is present. The discrimination evaluation index is mostly the area under the receiver operating characteristic curve (AUC, Area Under the Curve) or the C-index (consistency index) value. If the dependent variable of the model is a binary variable, the C-index value is the same as the AUC. Specifically, the receiver operating characteristic curve (ROC, Receiver Operating Characteristic) is drawn, and the AUC is calculated, and finally the model cutoff value is determined by the Youden index. Further, the Bootstrap self-sampling method, i.e., 1000 repeated samplings, is used to internally verify the model.

[0091] Among them, the calibration of the model is to evaluate whether the actually observed probability and the predicted probability are consistent, and the Hosmer-Lemeshow goodness of fit test and the method of drawing a calibration curve are often used. Statistical indicators are usually expressed as Brier scores, which are generally considered to measure the overall effect of the model. Specifically, the Hosmer-Lemeshow goodness of fit test is used for evaluation. Furthermore, a calibration curve is drawn, and the Bootstrap self-sampling method, that is, 1000 repeated samplings, is used to internally verify the model.

[0092] For example, the clinical benefit of the model is determined by drawing a clinical decision curve analysis (DCA), which integrates the preferences of patients and decision makers and can meet the actual needs of clinical decision-making.

[0093] In this embodiment, the method of evaluating and verifying the bone mass reduction risk prediction model and the osteoporosis risk prediction model is clarified to further ensure the discrimination, calibration and clinical benefit of the model.

[0094] Each module in the above-mentioned bone loss prediction system and bone loss prediction model establishment system can be implemented in whole or in part by software, hardware and their combination. Each of the above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory in the computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0095] In a specific embodiment, 553 research subjects were divided into three groups, including 178 normal bone mass group, 261 osteopenia group, and 114 osteoporosis group, of which osteopenia accounted for 47.1% and osteoporosis accounted for 20.6%. The male / female composition of the three groups was 113 (63.5%) / 65 (36.5%), 142 (54.4%) / 119 (45.6%), and 40 (35.1%) / 74 (64.9%), respectively. The proportion of women increased with the decrease of bone mass, and the comparison between gender groups was statistically significant (p < 0.001). No abnormal values ​​were found in the data of this embodiment, and the missing data ratio was less than 20%. Multiple interpolation method was used for interpolation, and a total of 20 interpolations and 50 iterations were performed, and finally the 20 data sets obtained were merged and summarized. The comparison of bone density of lumbar spine, femoral neck and hip among the three groups of patients showed that the p values ​​were all less than 0.05, and the differences were statistically significant, indicating that they were comparable. The comparison of general information, DXA examination results and laboratory data of the three groups of patients is shown in Table 1.

[0096] Table 1 (M[Q1, Q3] / n(%) / )

[0097]

[0098] Table 1

[0099]

[0100]

[0101] Table 1

[0102]

[0103]

[0104] Furthermore, in this embodiment, 36 possible risk factors are included, and whether or not osteopenia and osteoporosis occur is used as the dependent variable Y, and the 36 possible risk factors are used as the independent variable X, and each variable is assigned a value. The assignment of each variable is shown in Table 2.

[0105] Table 2

[0106]

[0107]

[0108] Since there may be multicollinearity among various variables, the LASSO regression with 10-fold cross validation was used to preliminarily screen the research variables, and the LASSO regression variable screening diagram was drawn, where Figure 3 (a) is the LASSO regression variable shrinkage coefficient diagram of the osteopenia risk prediction model, Figure 3 (b) is the relationship diagram between the LASSO regression 10-fold cross validation model error and Log (λ) of the osteopenia risk prediction model, Figure 4 (a) is the LASSO regression variable shrinkage coefficient diagram of the osteoporosis risk prediction model, and Figure 4 (b) is the relationship diagram between the LASSO regression 10-fold cross validation model error and Log (λ) of the osteoporosis risk prediction model. When the penalty coefficient λ value gradually increases, the variable coefficient will gradually be compressed until it increases to lambda.min, which is considered to be the best predictor screened out. When λ continues to increase to 1 standard error (lambda.1se), it is considered that the model performance is excellent and the number of independent variables is the least. In this embodiment, the predictive factors screened out when lambda.1se was selected, and the results of the LASSO regression of the final bone mass reduction risk prediction model showed that 10 predictive factors were screened out, namely, the potential predictive factor set for bone mass reduction, which were age, weight, systolic blood pressure, diabetes duration, blood creatinine, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, β-CTX, 25(OH)VD and parathyroid hormone; the results of the LASSO regression of the osteoporosis risk prediction model showed that 10 predictive factors were screened out, namely, the potential predictive factor set for osteoporosis, which were age, sex, weight, smoking history, drinking history, fasting insulin, β-CTX, TP1NP, 25(OH)VD and hemoglobin.

[0109] Furthermore, the variables screened by LASSO regression were included in the multivariate logistic regression analysis to determine the predictors of osteopenia in patients with T2DM. The results of multivariate logistic regression analysis (osteopenia risk prediction model) of osteopenia in patients with T2DM are shown in Table 3. Among them, age (OR = 1.052, 95% CI: 1.015-1.090, P = 0.005), body weight (OR = 0.967, 95% CI: 0.947-0.987, P = 0.001), serum creatinine (OR = 0.989, 95% CI: 0.979-0.999, P = 0.037), high-density lipoprotein cholesterol (OR = 2.688, 95% CI: 1.161-6.221, P = 0.02 1), β-CTX (OR=7.357, 95%CI: 2.559-21.149, P<0.001), 25(OH)VD (OR=0.966, 95%CI: 0.934-0.999, P=0.041) and parathyroid hormone (OR=1.016, 95%CI: 1.002-1.029, P=0.023) are all predictors of osteopenia in patients with T2DM (osteopenia risk prediction model). Among them, β represents the standardized regression coefficient, SE represents the standard error, Walds represents the chi-square value, P value is used to evaluate the effect of independent variables on dependent variables, OR value represents the odds ratio, OR95%CI represents the confidence interval of OR value, and constant represents the data related to but not influencing factors in the osteopenia risk prediction model.

[0110] Table 3

[0111]

[0112] Furthermore, the variables screened by LASSO regression were included in the multivariate logistic regression analysis to determine the predictors of osteoporosis in patients with T2DM. The results of multivariate logistic regression analysis (osteoporosis risk prediction model) of osteoporosis in patients with T2DM are shown in Table 4. Among them, age (OR = 1.067, 95% CI: 1.030-1.105, P < 0.001), female (OR = 6.368, 95% CI: 2.341-17.324, P < 0.001), weight (OR = 0.917, 95% CI: 0.889-0.945, P < 0.001), smoking history (OR = 7.643, 95% CI: 2.863-20.407, P < 0.001) and osteoporosis risk prediction model were significantly higher in patients with T2DM. ), drinking history (OR=6.431, 95%CI:2.695-15.342, P<0.001), TP1NP (OR=1.023, 95%CI:1.006-1.040, P=0.007) and 25(OH)VD (OR=0.946, 95%CI:0.905-0.988, P=0.012) are all predictors of osteoporosis in patients with T2DM (osteoporosis risk prediction model).

[0113] Table 4

[0114]

[0115] Furthermore, according to the results of multivariate logistic regression analysis, Figure 5 and Figure 6 As shown in the figure, R software was used to construct the nomogram models for T2DM patients with osteopenia and T2DM patients with osteoporosis, respectively. Figure 5 A nomogram model for T2DM patients with osteopenia (i.e., osteopenia risk prediction model). Figure 6 This is a nomogram model for T2DM patients with osteoporosis (i.e., osteoporosis risk prediction model). The score of each item in the prediction model is obtained by the score corresponding to the vertical line from each item in the nomogram to the scoring scale (Points), and then the score of each item is added to get the total score (TotalPoints). The value corresponding to the vertical line from the total score to the probability scale is the predicted probability of T2DM patients developing osteopenia or osteoporosis. Among them, vitamin D is 25(OH)VD.

[0116] For example, a patient is 55 years old, weighs 75 kg, has blood creatinine of 160 umol / L, high-density lipoprotein cholesterol of 1.2 mmol / L, β-CTX of 0.8 ng / mL, 25(OH)VD of 30 ng / mL, and parathyroid hormone of 60 pg / mL. The osteopenia nomogram (osteopenia risk prediction model) can be used to obtain the points of 10, 50, 35, 25, 50, 20, and 30, respectively. The total score is 220, and the corresponding probability is 0.38. For example, a patient aged 60 years old, female, weighing 75 kg, with no smoking history, a drinking history, TP1NP90 ng / mL, and 25(OH)VD35 ng / mL can use the osteoporosis nomogram (osteoporosis risk prediction model) to obtain the points: 10 points, 30 points, 50 points, 0 points, 30 points, 30 points and 10 points respectively. The total score is 160 points, and the corresponding probability is 0.5.

[0117] Furthermore, based on the results of multivariate logistic regression analysis, web calculators for two models were generated using R software. Web calculator for bone loss risk prediction model Figure 7 As shown in the following figure, the osteoporosis risk prediction model web calculator is as follows Figure 8 As shown, when using, input the patient's relevant information and click "Predict" to obtain the probability of osteopenia or osteoporosis in the T2DM patient and the 95% CI.

[0118] Furthermore, the receiver operating characteristic curve (ROC) was drawn, and the area under the curve (AUC) was used to evaluate the model's discriminatory ability. As shown in Figure 9(a), the AUC of the osteopenia risk prediction model was 0.730 (95% CI: 0.682-0.777), and its cutoff value was 0.611 when the Youden index was the largest, with a specificity of 71.9% and a sensitivity of 65.9%. Further calculations showed that the cut-off of the total score of the nomogram was 253 points and the corresponding probability was 61%, that is, patients with a total score of 253 points (probability 61%) or above were at high risk of T2DM complicated with osteopenia; as shown in the figure As shown in Figure 9(b), the AUC of the osteoporosis risk prediction model was 0.806 (95% CI: 0.760-0.852). When the Youden index was the largest, the cutoff value was 0.284, the specificity was 67.4%, and the sensitivity was 85.1%. Further calculation showed that the cut-off of the total score of the nomogram was 145.6 points and the corresponding probability was 28%, that is, patients with a total score of 145.6 points (probability 28%) or above were at high risk of T2DM complicated by osteoporosis.

[0119] Furthermore, the Bootstrap method was used for internal validation with 1000 self-sampling. As shown in Figure 10(a), the C-index of the bone mass reduction risk prediction model was 0.730 (95% CI: 0.728-0.731), and as shown in Figure 10(b), the C-index of the osteoporosis risk prediction model was 0.806 (95% CI: 0.804-0.807).

[0120] Further, the ROC comparison diagram of the nomogram of the risk prediction model for osteopenia and a single risk factor is shown in Figure 11 (a), and the ROC comparison diagram of the nomogram of the risk prediction model for osteoporosis and a single risk factor is shown in Figure 11 (b). Compared with single risk factors (such as bone metabolism markers: β-CTX, TP1NP), the nomogram model has the largest AUC, where age represents age, BCTX represents β-collagen special sequence, Cr represents blood creatinine, HDLC represents high-density lipoprotein cholesterol, Nomogram represents nomogram, PTH represents parathyroid hormone, VD represents 25-hydroxyvitamin D, weight represents body weight, drinking represents drinking history, TP1NP represents total type I procollagen amino-terminal extension peptide, sex represents gender, and smoking represents smoking history. Therefore, the two models in this embodiment have a good ability to distinguish between predicting the occurrence of osteopenia or osteoporosis in T2DM patients.

[0121] Furthermore, the goodness of fit of the two models was tested using the Homers-Lemeshow (HL) test. The HL test likelihood ratio chi-square value of the osteopenia risk prediction model was 6.367, and the P value was 0.703 (P>0.05); the HL test likelihood ratio chi-square value of the osteoporosis risk prediction model was 12.572, and the P value was 0.183 (P>0.05), indicating that there was no significant difference between the predicted probabilities and the true values ​​of the two models.

[0122] Furthermore, the calibration curves of the two prediction models were drawn using R software, and internal validation was performed by 1000 bootstrap samplings. The internal validation of the calibration line and calibration degree of the osteopenia risk prediction model is shown in Figure 12(a), and the internal validation of the calibration line and calibration degree of the osteoporosis risk prediction model is shown in Figure 12(b), where the dotted line (Ideal) represents the ideal reference line, the blue line (Apparent) represents the calibration curve of the prediction model, and the red line (Bias-corrected) represents the calibration curve after Bootstrap sampling. The closer the model calibration curve is to the baseline, the better the calibration of the model. The Brier scores of the osteopenia risk prediction model and the osteoporosis risk prediction model were 0.204 and 0.160, respectively, both less than 0.25, indicating that the two models had good overall effects. The overall effect score refers to the sum of the overall effect scores of discrimination and calibration. Therefore, the internal validation of the HL test and calibration curves indicated that the calibration of the two models was good.

[0123] Furthermore, the clinical decision curve (DCA) of the nomogram risk model was drawn using R software and compared with the single risk factor in the nomogram. The DCA curve of the nomogram of the osteopenia risk prediction model and the single risk factor is shown in Figure 13 (a), and the DCA curve of the nomogram of the osteoporosis risk prediction model and the single risk factor is shown in Figure 13 (b). Age represents age, sex represents gender, weight represents weight, drinking represents drinking history, smoking represents smoking history, Cr represents serum creatinine, HDLC represents high-density lipoprotein cholesterol, PTH represents parathyroid hormone, VD represents 25-hydroxyvitamin D, BCTX represents β-collagen special sequence, TP1NP represents total type I procollagen amino terminal extension peptide, and Nomogram represents nomogram. Among them, the horizontal axis represents the probability threshold, the vertical axis represents the net benefit rate, the blue line (None) means that no one intervenes, and the clinical net benefit is 0; the orange line (All) means that everyone intervenes, and the benefit rate is a negative backslash. The decision curve of the nomogram is above and to the right of the two ineffective lines, indicating that the model has a certain clinical predictive value and that the nomogram model has higher clinical benefit than a single risk factor.

[0124] Based on the same inventive concept, the embodiments of the present application also provide a bone loss prediction method using the above-mentioned bone loss prediction system and a bone loss prediction model establishment method using the above-mentioned bone loss prediction model establishment system. The implementation solutions to the problems provided by the above-mentioned methods are similar to the implementation solutions recorded in the above-mentioned systems. Therefore, the specific limitations in the embodiments of one or more bone loss prediction methods and bone loss prediction model establishment methods provided below can refer to the limitations on the bone loss prediction system and the bone loss prediction model establishment system in the above text, and will not be repeated here.

[0125] In an exemplary embodiment, Fig.14 As shown, a method for predicting bone loss is provided, the method comprising the following steps:

[0126] Step 1402: Obtain a reference influencing factor data set of the object to be tested.

[0127] Step 1404, calling a pre-established bone loss prediction model, inputting a reference influencing factor data set of the subject to be tested, and obtaining a bone loss prediction result of the subject to be tested.

[0128] In an exemplary embodiment, Fig.15 As shown, a method for establishing a bone loss prediction model is provided, which is used to establish the bone loss prediction model provided in the above embodiment, comprising:

[0129] Step 1502, obtain a preset number of bone mass data, and the influencing factor data sets corresponding to each bone mass data, and know the normal bone mass group, osteopenia group and osteoporosis group into which the bone mass data are divided according to the DXA examination results, obtain each bone mass reduction probability and its corresponding influencing factor data set, and form a single reduction sample with a single bone mass reduction probability and its corresponding influencing factor data set to obtain a reduction sample set; obtain each osteoporosis probability and its corresponding influencing factor data set, and form a single loose sample with a single osteoporosis probability and its corresponding influencing factor data set to obtain a loose sample set.

[0130] Step 1504, using the dataset of influencing factors in the reduction sample as input and the probability of osteopenia in the reduction sample as output, train to obtain an osteopenia risk prediction model; using the dataset of influencing factors in the loose sample as input and the probability of osteoporosis in the loose sample as output, train to obtain an osteoporosis risk prediction model.

[0131] Step 1506, generating a bone loss visualization display tool based on the osteopenia risk prediction model and the osteoporosis risk prediction model, and evaluating and verifying the osteopenia risk prediction model and the osteoporosis risk prediction model, to obtain a bone loss prediction model with a reference influencing factor data set as input and a bone loss prediction result as output; wherein the bone loss prediction model includes the evaluated and verified osteopenia risk prediction model and osteoporosis risk prediction model.

[0132] In yet another embodiment, Fig.16 As shown in the figure, a method for establishing a bone mass loss prediction model is provided. First, the final inclusion subjects, i.e., the research subjects of the bone mass loss prediction model, are determined according to the inclusion and exclusion criteria; data are collected through the hospital electronic medical record system, including general data, clinical data and DXA examination results of T2DM patients hospitalized in the endocrinology department, i.e., bone mass data and a data set of influencing factors corresponding to each bone mass data; and then the research subjects are grouped according to the DXA results, into a normal bone mass group, a bone loss group, and an osteoporosis group. LASSO regression was used to screen variables, and a potential predictive factor set of osteopenia was obtained based on the normal bone mass group and the osteopenia group. The potential predictive factor set of osteopenia was included in the Logistic multivariate analysis to generate the first model T2DM combined with osteopenia risk prediction model, namely the osteopenia risk prediction model; LASSO regression was used to screen variables based on the osteopenia group and the osteoporosis group to obtain a potential predictive factor set of osteoporosis, and the potential predictive factor set of osteoporosis was included in the Logistic multivariate analysis to generate the second model T2DM combined with osteoporosis risk prediction model, namely the osteoporosis risk prediction model, and the corresponding nomograms and web calculators were generated based on the osteopenia risk prediction model and the osteoporosis risk prediction model. Furthermore, the osteopenia risk prediction model and the osteoporosis risk prediction model were evaluated for discrimination, calibration and clinical benefit, and then the Bootstrap method was used for verification to obtain the bone loss prediction model; wherein, the bone loss prediction model includes the osteopenia risk prediction model and the osteoporosis risk prediction model after evaluation and verification.

[0133] In another embodiment, Fig.17 As shown, a method for predicting bone loss is provided, comprising:

[0134] Step 1702, obtain a preset number of bone mass data, and the influencing factor data sets corresponding to each bone mass data, and know the normal bone mass group, osteopenia group and osteoporosis group into which the bone mass data are divided according to the DXA examination results, obtain each bone mass reduction probability and its corresponding influencing factor data set, and form a single reduction sample with a single bone mass reduction probability and its corresponding influencing factor data set to obtain a reduction sample set; obtain each osteoporosis probability and its corresponding influencing factor data set, and form a single loose sample with a single osteoporosis probability and its corresponding influencing factor data set to obtain a loose sample set.

[0135] Step 1704, taking the influencing factor dataset in the reduced sample as an independent variable, using LASSO regression to screen the influencing factor dataset in the reduced sample according to the normal bone mass group and the osteopenia group, and obtaining a set of potential predictive factors for osteopenia.

[0136] Step 1706, incorporating the potential predictive factor set of osteopenia into the Logistic regression analysis, obtaining the predictive factor set of osteopenia, and generating an osteopenia risk prediction model.

[0137] Step 1708, taking the influencing factor dataset in the osteoporosis sample as an independent variable, using LASSO regression to screen the influencing factor dataset in the osteoporosis sample according to the osteopenia group and the osteoporosis group, and obtaining a set of potential predictive factors for osteoporosis.

[0138] Step 1710, incorporating the osteoporosis potential prediction factor set into the Logistic regression analysis, obtaining the osteoporosis prediction factor set, and generating an osteoporosis risk prediction model.

[0139] Step 1712, generating an osteopenia risk prediction nomogram and web calculator, an osteoporosis risk prediction nomogram and web calculator based on the osteopenia risk prediction model and the osteoporosis risk prediction model through the target software.

[0140] Step 1714, the discrimination of the osteopenia risk prediction model and the osteoporosis risk prediction model is evaluated by using the receiver operating characteristic curve and verified by using the Bootstrap method.

[0141] Step 1716, the calibration of the osteopenia risk prediction model and the osteoporosis risk prediction model is evaluated by Homers-Lemeshow test and verified by Bootstrap method.

[0142] Step 1718, evaluating the clinical benefits of the osteopenia risk prediction model and the osteoporosis risk prediction model through a clinical decision curve diagram.

[0143] Step 1720, obtaining a bone loss prediction model with the reference influencing factor data set as input and the bone loss prediction result as output; wherein the bone loss prediction model includes an osteopenia risk prediction model and an osteoporosis risk prediction model that have been evaluated and verified.

[0144] Step 1722, obtaining a reference influencing factor data set of the object to be tested.

[0145] Step 1724, calling a pre-established bone loss prediction model, inputting a reference influencing factor data set of the subject to be tested, and obtaining a bone loss prediction result of the subject to be tested.

[0146] It should be understood that, although the steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence as indicated by the arrows, they are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps. At least a part of the steps in the flowcharts involved in the above-mentioned embodiments may include multiple steps or multiple stages, which may be executed at different times, and the execution order of these steps or stages may be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0147] In one embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the functions of the above-mentioned system embodiments when executing the computer program.

[0148] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the functions of the above-mentioned system embodiments are implemented.

[0149] In one embodiment, a computer program product is provided, including a computer program, which implements the functions of the above-mentioned system embodiments when executed by a processor.

[0150] The technical features of the above embodiments may be combined arbitrarily, and as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0151] The above-described embodiments only express several implementation methods of the present application, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for those of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A bone loss prediction system, characterized in that: The system comprises: A reference data acquisition module is used to obtain a reference influencing factor data set of the object to be tested; The result data output module is used to call the pre-established bone loss prediction model, input the reference influencing factor data set of the object to be tested, and obtain the bone loss prediction result of the object to be tested.

2. A bone loss prediction model establishment system, used to establish the bone loss prediction model as claimed in claim 1, characterized in that: The system comprises: The sample data acquisition and grouping module is used to obtain a preset number of bone mass data and the influencing factor data sets corresponding to each bone mass data, and to know the normal bone mass group, osteopenia group and osteoporosis group into which the bone mass data are divided according to the DXA examination results, to obtain each osteopenia probability and its corresponding influencing factor data set, and to form a single reduction sample with a single osteopenia probability and its corresponding influencing factor data set to obtain a reduction sample set; to obtain each osteoporosis probability and its corresponding influencing factor data set, and to form a single loose sample with a single osteoporosis probability and its corresponding influencing factor data set to obtain a loose sample set; A model training module is used to train a risk prediction model for osteopenia by taking a dataset of influencing factors in a reduction sample as input and a probability of osteopenia in a reduction sample as output; and to train a risk prediction model for osteoporosis by taking a dataset of influencing factors in a loose sample as input and a probability of osteoporosis in a loose sample as output; An evaluation and verification module is used to generate a bone loss visualization display tool based on the bone mass reduction risk prediction model and the osteoporosis risk prediction model, and to evaluate and verify the bone mass reduction risk prediction model and the osteoporosis risk prediction model to obtain a bone mass loss prediction model with a reference influencing factor data set as input and a bone mass loss prediction result as output; wherein the bone mass loss prediction model includes the evaluated and verified bone mass reduction risk prediction model and the osteoporosis risk prediction model.

3. The system according to claim 2, characterized in that The method of training a risk prediction model for osteopenia by reducing the influencing factor data set in the sample as input and reducing the probability of osteopenia in the sample as output comprises: The influencing factor dataset in the reduced sample is used as an independent variable, and LASSO regression is used to screen the influencing factor dataset in the reduced sample according to the normal bone mass group and the osteopenia group to obtain a set of potential predictive factors for osteopenia; The potential predictive factor set for osteopenia was included in the logistic regression analysis to obtain the predictive factor set for osteopenia and generate an osteopenia risk prediction model.

4. The system according to claim 2, characterized in that The method of taking the influencing factor data set in the osteoporosis sample as input and the osteoporosis probability in the osteoporosis sample as output, and training to obtain the osteoporosis risk prediction model includes: Taking the influencing factor data set in the osteoporosis sample as an independent variable, using LASSO regression to screen the influencing factor data set in the osteoporosis sample according to the osteopenia group and the osteoporosis group, to obtain a set of potential predictive factors for osteoporosis; The osteoporosis potential predictive factor set is incorporated into the Logistic regression analysis to obtain the osteoporosis predictive factor set and generate an osteoporosis risk prediction model.

5. The system according to claim 2, characterized in that The bone loss visualization display tool generated according to the osteopenia risk prediction model and the osteoporosis risk prediction model comprises: The target software generates an osteopenia risk prediction nomogram and a web calculator, an osteoporosis risk prediction nomogram and a web calculator based on the osteopenia risk prediction model and the osteoporosis risk prediction model.

6. The system according to claim 2, characterized in that The evaluation and verification of the osteopenia risk prediction model and the osteoporosis risk prediction model includes: The osteopenia risk prediction model and osteoporosis risk prediction model were evaluated for discrimination, calibration, and clinical benefit by designated methods, and verified by the Bootstrap method.

7. The system according to claim 6, characterized in that The evaluation of the differentiation, calibration and clinical benefit of the osteopenia risk prediction model and the osteoporosis risk prediction model by a specified method includes: The receiver operating characteristic curve was used to evaluate the discrimination between the osteopenia risk prediction model and the osteoporosis risk prediction model. The Homers-Lemeshow test was used to evaluate the calibration of the osteopenia risk prediction model and the osteoporosis risk prediction model; The clinical decision curve was used to evaluate the clinical benefits of the osteopenia risk prediction model and osteoporosis risk prediction model.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the functions of the system according to any one of claims 1 to 7 are realized.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the functions of the system according to any one of claims 1 to 7 are realized.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the functions of the system according to any one of claims 1 to 7 are realized.

Images