Construction and application method and device of chronic kidney disease patient vulnerability risk prediction model
By constructing a vulnerability risk prediction model for patients with chronic kidney disease, and using machine learning models and recursive feature elimination to identify key features, this approach addresses the lack of vulnerability risk assessment for patients with chronic kidney disease in existing technologies, achieving efficient risk prediction and interpretability.
Patent Information
- Application Number
- CN202511557633.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-10-29
- Publication Date
- 2026-01-23
AI Technical Summary
There is currently no vulnerability risk prediction model for patients with chronic kidney disease, making it impossible to accurately assess their risk.
A vulnerability risk prediction model for chronic kidney disease patients based on a publicly available sample set was constructed. By acquiring the basic characteristics of the subjects, a machine learning model was trained and a recursive feature elimination method was applied to identify key features, and a target machine learning model was established for prediction.
It enables accurate prediction of vulnerability risk in patients with chronic kidney disease, improves the performance and interpretability of the prediction model, and provides clinical application value.
Smart Images
Figure CN121393885A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of data processing, in particular to a construction method and device of a frailty risk prediction model for patients with chronic kidney disease. BACKGROUND
[0002] In the past, studies on different populations have established frailty prediction models. For example, some researchers have constructed a nomogram model based on waist circumference, cognitive function, grip strength, social activity and depression status in the diabetic population, with an AUC of 0.912; some researchers have established a frailty risk prediction model based on age, professional experience, economic status and oral health in the community elderly population, with an AUC of 0.881; and another study has established a related model in elderly patients with coronary heart disease. However, there has been no report on a frailty risk prediction model for patients with chronic kidney disease (CKD). SUMMARY
[0003] The purpose of the present application is to provide a construction method and device of a frailty risk prediction model for patients with chronic kidney disease, which can accurately predict the frailty risk of patients with chronic kidney disease.
[0004] To achieve the above-mentioned purpose, the present application provides the following solutions: In a first aspect, the present application provides a construction method of a frailty risk prediction model for patients with chronic kidney disease, comprising: obtaining a public sample set; the public sample set comprising basic characteristics and corresponding frailty state results of a plurality of subjects; the subjects including patients with chronic kidney disease; training a plurality of machine learning models based on the public sample set, determining a machine learning model with optimal model performance, and obtaining a best machine learning model; determining key features in the basic characteristics of the subjects based on the best machine learning model using a recursive feature elimination method; retraining the best machine learning model based on the key features and the corresponding frailty state results to obtain a target machine learning model; the target machine learning model is used for predicting the frailty risk of patients with chronic kidney disease.
[0005] In a second aspect, the present application provides an application method of a frailty risk prediction model for patients with chronic kidney disease, comprising: obtaining key features of a patient with chronic kidney disease to be predicted; inputting the key features of the patient with chronic kidney disease to be predicted into a target machine learning model to obtain a corresponding frailty risk prediction result; the target machine learning model is constructed based on the construction method of the frailty risk prediction model for patients with chronic kidney disease described above.
[0006] In a third aspect, the present application provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor executes the computer program to implement the method for constructing a chronic kidney disease patient vulnerability risk prediction model or the method for applying the chronic kidney disease patient vulnerability risk prediction model.
[0007] In a fourth aspect, the present application provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the method for constructing a chronic kidney disease patient vulnerability risk prediction model or the method for applying the chronic kidney disease patient vulnerability risk prediction model.
[0008] In a fifth aspect, the present application provides a computer program product, comprising a computer program, and the computer program is executed by a processor to implement the method for constructing a chronic kidney disease patient vulnerability risk prediction model or the method for applying the chronic kidney disease patient vulnerability risk prediction model.
[0009] According to the specific embodiments provided by the present application, the present application discloses the following technical effects: The present application provides a method and device for constructing and applying a chronic kidney disease patient vulnerability risk prediction model, the method for constructing the model comprising obtaining a public sample set; the public sample set comprising basic characteristics and corresponding vulnerability state results of a plurality of subjects; the subjects including chronic kidney disease patients; training a plurality of machine learning models based on the public sample set, determining a machine learning model with optimal model performance, and obtaining a best machine learning model; determining key features in the basic characteristics of the subjects based on the best machine learning model and using a recursive feature elimination method; retraining the best machine learning model based on the key features and the corresponding vulnerability state results, and obtaining a target machine learning model; the target machine learning model being used for predicting the vulnerability risk of chronic kidney disease patients. The present application realizes accurate prediction of the vulnerability risk of chronic kidney disease patients. BRIEF DESCRIPTION OF DRAWINGS
[0010] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.
[0011] Figure 1 FIG. 1 is a schematic diagram of an application environment of a method for constructing a chronic kidney disease patient vulnerability risk prediction model according to an embodiment of the present application; Figure 2A flowchart of a method for constructing a chronic kidney disease patient vulnerability risk prediction model is provided in an embodiment of the present application. Figure 3 A technical concept diagram of a method for constructing a chronic kidney disease patient vulnerability risk prediction model is provided in an embodiment of the present application. Figure 4 ROC curve diagrams of six machine learning models generated based on a training data set are provided in an embodiment of the present application. Figure 5 ROC curve diagrams of six machine learning models generated based on an internal validation data set are provided in an embodiment of the present application. Figure 6 ROC curve diagrams of six machine learning models generated based on an external validation data set are provided in an embodiment of the present application. Figure 7 XGBoost model ROC curve diagrams with different numbers of features are provided in an embodiment of the present application. Figures 8-9 P-R curves of 50, 20, 15, 12, and 11 features are provided in an embodiment of the present application. Figure 10 A prediction performance diagram of an XGBoost model in an external validation queue is provided in an embodiment of the present application. Figure 11 A SHAP method explanation diagram of a global model is provided in an embodiment of the present application. Figure 12 and Figure 13 A SHAP method explanation diagram of a local model is provided in an embodiment of the present application. Figure 14 A flowchart of an application method of a chronic kidney disease patient vulnerability risk prediction model is provided in an embodiment of the present application. Figure 15 A structural diagram of a computer device is provided in an embodiment of the present application. DETAILED DESCRIPTION
[0012] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of the present application.
[0013] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application will be further described in detail below with reference to the drawings and specific embodiments.
[0014] The method for constructing a chronic kidney disease patient vulnerability risk prediction model provided in the embodiments of the present application can be applied in an application environment as shown in Figure 1 The terminal communicates with the server through a network. The data storage system can store data required to be processed by the server. The data storage system can be separately arranged, integrated on the server, placed on a cloud or other server. The terminal can send an open sample set (the open sample set includes basic features of a plurality of subjects and corresponding vulnerability state results; the subjects include chronic kidney disease patients) to the server. After receiving the open sample set, the server trains a plurality of machine learning models based on the open sample set, determines a machine learning model with optimal model performance, and obtains a best machine learning model. Based on the best machine learning model, a recursive feature elimination method is used to determine key features in the basic features of the subject. Based on the key features and the corresponding vulnerability state results, the best machine learning model is retrained to obtain a target machine learning model. The target machine learning model is used for predicting the vulnerability risk of chronic kidney disease patients. The server can feed back the obtained target machine learning model to the terminal. In addition, in some embodiments, the method for constructing a chronic kidney disease patient vulnerability risk prediction model can also be implemented by the server or the terminal alone, for example, the terminal can directly construct a chronic kidney disease patient vulnerability risk prediction model for the open sample set, or the server can obtain the open sample set from the data storage system and construct a chronic kidney disease patient vulnerability risk prediction model.
[0015] The terminal can be, but is not limited to, various desktop computers, notebook computers, smart phones, tablet computers, Internet of Things devices and portable wearable devices. The server can be implemented by an independent server or a server cluster composed of multiple servers, and can also be a cloud server.
[0016] In an exemplary embodiment, as shown in Figure 2 and Figure 3 A method for constructing a chronic kidney disease patient vulnerability risk prediction model is provided. The method is executed by a computer device, specifically by a terminal or a server or the like, and can be executed by the terminal and the server together. In the embodiments of the present application, the method is applied to the server in Figure 1 The method includes the following steps S100 to S400.
[0017] S100: Obtain an open sample set; the open sample set includes basic features of a plurality of subjects and corresponding vulnerability state results; the subjects include chronic kidney disease patients.
[0018] The present application is based on the National Health and Nutrition Examination Survey (NHANES) data from 1999-2020, which is conducted by the National Center for Health Statistics (NCHS) in collaboration with the Centers for Disease Control and Prevention (CDC) to assess the health and nutrition status of adults and children in the United States. NHANES data covers demographic characteristics, physical examination, body composition, comorbidities, clinical and laboratory tests, drug use, personal interviews, and related measurements, which can be used to estimate the prevalence of various acute and chronic diseases (diagnosed and undiagnosed), as well as health indicators such as nutritional status and chemical exposure. The database is publicly available for global researchers to download and apply. A total of 11,463 subjects were included, and randomly divided into training set, internal validation set and external validation set according to 70%:20%:10%. Among them, 7,979 people were assigned to the training set, 2,337 to the internal validation set, and 1,147 to the external validation set. There was no statistical difference in demographic and clinical characteristics among the three groups (see Table 1). The prevalence of frailty in the training set was 46.7% (3,724 / 7,979), which was close to that in the internal validation set (46.1%, 1,078 / 2,337) and the external validation set (46.6%, 534 / 1,147).
[0019] Table 1 Comparison of demographic and clinical characteristics between training, internal validation and external validation cohorts.
[0020] Continuous numerical values are expressed as median [interquartile range]. Categorical data are expressed as number (percentage). P<0.05 is statistically significant. eGFR: estimated glomerular filtration rate; ALP: alkaline phosphatase; LDH: lactate dehydrogenase.
[0021] To reduce the impact of multicollinearity on the prediction accuracy of the model, only one variable (feature) was retained when the Spearman correlation coefficient was >0.85. In addition, variables with a missing rate of more than 25% were excluded. Finally, a total of 51 variables were included for modeling, including 8 demographic characteristics, 14 comorbidities, 2 vital signs, 2 body composition indicators, 22 laboratory indicators, and 3 medication information. All 51 features were used to construct the prediction model. Missing data was handled by median imputation.
[0022] Definition of frailty: includes the following five indicators: ① unintentional weight loss; ② subjective fatigue; ③ decreased muscle strength; ④ slow walking speed; ⑤ reduced physical activity. If ≥3 items are met, it is determined as frailty.
[0023] S200: training a plurality of machine learning models based on the public sample set, determining a machine learning model with optimal model performance, and obtaining a best machine learning model.
[0024] S300: determining key features in the basic features of the subject based on the optimal machine learning model by applying recursive feature elimination.
[0025] S400: retraining the optimal machine learning model based on the key features and the corresponding frailty status results to obtain a target machine learning model; the target machine learning model is used for predicting the frailty risk of chronic kidney disease patients.
[0026] Implementing the above steps S100 to S400, first, based on the public sample set, a plurality of machine learning models are trained to obtain an optimal machine learning model, then based on the optimal machine learning model, recursive feature elimination is applied to determine the key features in the basic features of the subject, thereby obtaining a target machine learning model, and then predicting the frailty risk of chronic kidney disease patients according to the target machine learning model and the key features, achieving accurate prediction.
[0027] In another exemplary embodiment of the present application, in step S200, based on the public sample set, a plurality of machine learning models are trained to determine the machine learning model with the best model performance, and an optimal machine learning model is obtained, which specifically includes: (1) Based on the public sample set, a plurality of machine learning models are trained to obtain a plurality of trained machine learning models.
[0028] As an example, a total of 6 machine learning models are compared: support vector machine (SVM), extreme gradient boosting (XGBoost), random forest (RF), K-nearest neighbor (KNN), decision tree (DT) and logistic regression (LR).
[0029] (2) Determine the model performance evaluation index of the plurality of trained machine learning models; the model performance evaluation index includes at least one of the AUC index of the ROC curve, the positive predictive value (PPV), the negative predictive value (NPV), the sensitivity, the specificity, the accuracy and the F1 value.
[0030] The model prediction ability is mainly evaluated by the area under the receiver operating characteristic curve (AUC). The optimal cutoff value is determined by the Youden index (sensitivity + specificity - 1).
[0031] As an example, the hyperparameters of the model are optimized by grid search and 10-fold cross-validation to obtain the best AUC index, and then the optimal machine learning model is determined.
[0032] (3) Determine the machine learning model with the best model performance according to the model performance evaluation index to obtain the optimal machine learning model.
[0033] Table 2 summarizes the performance of each model on the training set, internal and external validation set,Figures 4-6 The ROC curves are shown. Figure 4 ROC curves of the six machine learning models generated based on the training dataset, Figure 5 ROC curves of the six machine learning models generated based on the internal validation dataset, Figure 6 ROC curves of the six machine learning models generated based on the external validation dataset. The results show that the XGBoost model has the best discrimination ability: Internal validation set: AUC = 0.87 (95% CI: 0.86-0.88), accuracy 0.79, sensitivity 73%, specificity 83%, PPV 79%, NPV 79%; External validation set: AUC = 0.87 (95% CI: 0.85-0.89), sensitivity 74%, specificity 84%.
[0034] Given that XGBoost is superior to other models in various performance indicators, it is selected as the final model.
[0035] Table 2 Performance of each model in the training set, internal and external validation sets
[0036] In another exemplary embodiment of the present application, feature selection is crucial in machine learning, which can reduce dimensionality, improve interpretability and prediction performance, and reduce training time. In the best-performing model, the present application applies the recursive feature elimination (RFE) method to systematically compare the performance of models with different numbers of variables (e.g. 11-15, 20, 50), and uses DeLong test to compare the AUC of different models to determine the final optimal feature subset (key feature set). Therefore, in step S300, based on the best machine learning model, the recursive feature elimination method is used to determine the key features in the basic features of the subject, specifically including: (1) Determine different numbers of features using the recursive feature elimination method.
[0037] (2) Input each number of features corresponding to the basic features into the best machine learning model to obtain the model performance of each best machine learning model.
[0038] (3) Use DeLong test to compare the AUC indicators of the ROC curves of each best machine learning model.
[0039] (4) Determine the key features in the basic features according to the comparison results of the AUC indicators.
[0040] In the feature selection process of the XGBoost model, the XGBoost-RFE algorithm was used to gradually reduce the number of variables. Although the model containing 51 features was slightly higher than the 12-feature model in the area under the P-R curve (AUC: 0.87 vs 0.85, ΔAUC = 0.02), the simplified 12-feature model was more feasible and interpretable in clinical application. The final model included 12 features: cardiovascular disease, antihypertensive drug use, hemoglobin, hypoglycemic drug use, serum albumin, estimated glomerular filtration rate (eGFR), C-reactive protein, age, alkaline phosphatase, blood glucose, lactate dehydrogenase, and total cholesterol.
[0041] Figures 7-9 The prediction performance of the XGBoost model under different numbers of features is shown. Among them, Figure 7 The ROC curves of XGBoost models with different numbers of features are shown. Figures 8-9 The P-R curves correspond to 50, 20, 15, 12, and 11 features, respectively. Figures 7-9 The prediction effect of the XGBoost model in the internal validation queue is intuitively presented.
[0042] Figure 10 The prediction performance of the XGBoost model in the external validation queue is shown. Figure 10 Among them, A is the AUC curve, and B is the P-R curve. The AUC curve and the P-R curve show the performance of the XGBoost model in the external validation queue. AUC is the area under the receiver operating characteristic curve; P-R is the precision-recall rate.
[0043] The XGBoost model performs stably in predicting the fragile state of CKD patients: AUC = 0.85, sensitivity = 73%, specificity = 82%, PPV = 77%, NPV = 78%, and accuracy = 0.78. The calibration curve and the decision curve analysis (such as Figure 4 and Figure 7 ) both indicate the reliability of the prediction and the clinical value.
[0044] In another exemplary embodiment of the present application, machine learning models are often considered "black boxes" and it is difficult to explain their prediction basis. To improve interpretability, the present application uses the SHAP (Shapley additive explanation) method to reveal the key factors of model prediction from global and individual levels. SHAP has a solid theoretical basis and can assign stable and consistent contribution values to each variable, and show the relationship between them and the fragile state. Global explanation reveals the contribution of each variable (feature) to the overall model, and local explanation can give individualized prediction basis for a single patient. Based on the SHAP feature importance ranking, the feature with the best prediction ability is determined. Therefore, the method for constructing the chronic kidney disease patient fragility risk prediction model further comprises: (1) The SHAP method was used to evaluate the global and local contributions of each basic feature to the vulnerability risk prediction results.
[0045] (2) Important features are determined based on the global and local contributions of each basic feature; these important features are used to assist in determining key features. Generally, important features are basically the same as the aforementioned key features.
[0046] To provide a standardized and easily understood framework for model interpretation, this application uses the SHAP method to evaluate the contribution of each variable to the prediction results. Global interpretation: SHAP summary diagram ( Figure 11 The data in sections A and B show that, ranked by importance, the average contribution values of each variable are as follows: cardiovascular disease, eGFR, hemoglobin, C-reactive protein, and age are the main influencing factors. Figure 11 This section illustrates the interpretability of the SHAP method for the global model. A is a summary bar chart of SHAP data, and B is a summary scatter plot of SHAP data. The probability value corresponding to each feature in the model increases with its SHAP value. Each patient's SHAP value in the model corresponds to a scatter plot, thus each patient has a scatter plot representing that feature on the curve. The scatter plot color reflects the actual value of each patient's feature; yellow indicates a higher feature value, and purple indicates a lower feature value. The scatter plots are stacked vertically to show the density distribution. CVD: Cardiovascular disease; eGFR: Estimated glomerular filtration rate; CRP: C-reactive protein; ALP: Alkaline phosphatase; LDH: Lactate dehydrogenase; CHOL: Cholesterol.
[0047] Partial Explanation: SHAP Waterfall Chart ( Figure 12 A and B in the middle and Figure 13 C and D in the diagram illustrate the predictive pathway for a single CKD patient, and by combining individualized input data, reveal the formation mechanism of specific predictions. Figure 12 and Figure 13 This is the SHAP method's interpretation of the local model. Yellow represents decisions that increase vulnerability risk, while purple represents decisions that decrease vulnerability risk. Figure 12 and Figure 13 In the diagram, A and C, and B and D, correspond to the force plot and waterfall plot of a single sample, respectively, and the evolution of the risk contribution of each feature. Figure 12 A and Figure 13 C in the diagram illustrates the progression of an individual chronic kidney disease patient towards the "vulnerable" category. Figure 12 B and Figure 13 The "D" in the data represents the evolution of individual patients towards the "non-vulnerable" category. The results indicate that cardiovascular disease, decreased renal function, high cholesterol, low hemoglobin, and advanced age all significantly increase the vulnerability risk in CKD patients.
[0048] The application fills this gap for the vulnerability risk prediction model of CKD population. The XGBoost model based on 12 variables not only achieves high prediction accuracy and recall rate, but also provides visual interpretation through SHAP method and develops an online tool for clinical application.
[0049] In another exemplary embodiment of the application, the method for constructing the vulnerability risk prediction model of chronic kidney disease patients further comprises: The target machine learning model is evaluated by using a calibration curve, a precision-recall curve and a decision curve.
[0050] Among them, the calibration curve is used to evaluate the consistency of the predicted probability and the true outcome, the precision-recall (P-R) curve is used to verify the performance of the model in the class imbalance situation, the decision curve analysis (DCA) is used to evaluate the clinical application value, and bilateral P<0.05 is considered statistically significant. In addition, five-fold and ten-fold cross-validation are performed in the external validation set to further confirm the stability of the model.
[0051] In the external validation set, the AUC of the XGBoost model is 0.846, which is comparable to the performance of the internal validation set (ΔAUC=0.005, as shown in Figure 10 A), and the area under the precision-recall curve is 0.823 (as shown in Figure 10 B). The calibration curve and the decision curve (as shown in Figure 5 and Figure 8 ) further support its stability and clinical practicability.
[0052] In another exemplary embodiment of the application, as shown in Figure 14 , a method for applying the vulnerability risk prediction model of chronic kidney disease patients is provided, comprising: T100: Obtain the key features of the chronic kidney disease patient to be predicted.
[0053] T200: Input the key features of the chronic kidney disease patient to be predicted into the target machine learning model to obtain the corresponding vulnerability risk prediction result; the target machine learning model is constructed based on the method for constructing the vulnerability risk prediction model of chronic kidney disease patients described above.
[0054] In order to facilitate clinical promotion and application, the final target machine learning model can be developed into a web tool based on the Streamlit framework. Clinicians only need to input the relevant features (key features) of the patient to obtain the prediction probability of the occurrence of vulnerability state in real time, thereby assisting clinical decision-making.
[0055] The application further provides an application scenario of the application of the construction and application method of the chronic kidney disease patient vulnerability risk prediction model. Specifically, the construction and application method of the chronic kidney disease patient vulnerability risk prediction model provided in the embodiment can be applied in a chronic kidney disease patient vulnerability risk prediction scenario. The scenario includes a model construction link and a prediction link. The model construction link is used for constructing a chronic kidney disease patient vulnerability risk prediction model; and the prediction link is used for performing vulnerability risk prediction based on the constructed chronic kidney disease patient vulnerability risk prediction model. The construction method of the chronic kidney disease patient vulnerability risk prediction model provided in the embodiment belongs to the model construction link. The application method of the chronic kidney disease patient vulnerability risk prediction model provided in the embodiment belongs to the prediction link.
[0056] In an exemplary embodiment, a computer device is provided, which can be a server or a terminal, and an internal structure diagram thereof can be as shown in Figure 15 The computer device includes a processor, a memory, an input / output interface (I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store construction data of a chronic kidney disease patient vulnerability risk prediction model. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals through network connection. The computer program is executed by the processor to implement a construction method of a chronic kidney disease patient vulnerability risk prediction model or an application method of a chronic kidney disease patient vulnerability risk prediction model.
[0057] Those skilled in the art can understand that Figure 15 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. A specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components. In an exemplary embodiment, a computer device is provided, which includes a memory and a processor, and the memory stores a computer program. The processor executes the computer program to implement the steps in each method embodiment described above.
[0058] In an exemplary embodiment, a computer readable storage medium storing a computer program is provided, the computer program, when executed by a processor, implements the steps of any of the above method embodiments.
[0059] In an exemplary embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the steps of any of the above method embodiments.
[0060] It should be noted that the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties, and the collection, use and processing of related data need to comply with relevant regulations.
[0061] It can be understood by those skilled in the art that all or part of the processes in the above embodiments can be completed by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer readable storage medium. When the computer program is executed, it can include the processes of the above embodiments. Any reference to memory, database or other medium used in the embodiments provided by the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration but not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.
[0062] The database involved in each of the embodiments provided in the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a blockchain, and the like, without being limited thereto. The processor involved in each of the embodiments provided in the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, and the like, without being limited thereto.
[0063] Any combination of the technical features of the above embodiments can be made. In order to make the description simple, all possible combinations of the technical features in the above embodiments are not described, however, as long as the combination of the technical features does not exist, it should be considered as the scope of the present application.
[0064] The principles and implementation modes of the present application are described by applying specific examples herein, and the above embodiments are only used to help understand the method of the present application and its core idea; meanwhile, for those skilled in the art, according to the idea of the present application, the specific implementation mode and application range will be changed. In conclusion, the content of the present application should not be understood as a limitation.
Claims
1. A method of constructing a chronic kidney disease patient vulnerability risk prediction model, characterized by, The application comprises the following steps: obtaining a public sample set, wherein the public sample set comprises basic characteristics and corresponding vulnerability state results of a plurality of subjects, and the subjects include chronic kidney disease patients; training a plurality of machine learning models based on the public sample set, determining a machine learning model with optimal model performance, and obtaining an optimal machine learning model; based on the optimal machine learning model, applying a recursive feature elimination method to determine key features in the basic characteristics of the subjects; retraining the optimal machine learning model based on the key features and the corresponding vulnerability state results, and obtaining a target machine learning model; the target machine learning model is used for predicting the vulnerability risk of chronic kidney disease patients.
2. The method for constructing a vulnerability risk prediction model for patients with chronic kidney disease according to claim 1, characterized in that, The application comprises the following steps: training a plurality of machine learning models based on the public sample set, determining a machine learning model with optimal model performance, and obtaining an optimal machine learning model, specifically comprising the following steps: training a plurality of machine learning models based on the public sample set, and obtaining a plurality of trained machine learning models; determining model performance evaluation indexes of the plurality of trained machine learning models; the model performance evaluation indexes include at least one of AUC indexes of ROC curves, PPV indexes, NPV indexes, sensitivities, specificities, accuracies, and F1 values; 3. The method for constructing a vulnerability risk prediction model for patients with chronic kidney disease according to claim 1, characterized in that, determining a machine learning model with optimal model performance according to the model performance evaluation indexes, and obtaining an optimal machine learning model. Based on the optimal machine learning model, the application applies a recursive feature elimination method to determine key features in the basic characteristics of the subjects, specifically comprising the following steps: determining different feature quantities by using the recursive feature elimination method; inputting the basic characteristics corresponding to each feature quantity into the optimal machine learning model, and obtaining model performances of the optimal machine learning models; comparing AUC indexes of ROC curves of the optimal machine learning models by using DeLong test; 4. The method for constructing a vulnerability risk prediction model for patients with chronic kidney disease according to claim 1, characterized in that, determining key features in the basic characteristics according to the comparison results of the AUC indexes. The application further comprises the following steps: evaluating global contributions and local contributions of each basic characteristic to the vulnerability risk prediction results by using a SHAP method; 5. The method for constructing a vulnerability risk prediction model for patients with chronic kidney disease according to claim 1, characterized in that, determining important features according to the global contributions and the local contributions of each basic characteristic; the important features are used for assisting in determining the key features. The application further comprises the following steps:
6. The method for constructing a vulnerability risk prediction model for patients with chronic kidney disease according to claim 1, characterized in that, evaluating the target machine learning model by using a calibration curve, a precision-recall curve, and a decision curve.
7. A method of applying a chronic kidney disease patient vulnerability risk prediction model, characterized in that, The key features include cardiovascular diseases, antihypertensive drug use, hemoglobin, antidiabetic drug use, serum albumin, estimated glomerular filtration rate, C-reactive protein, age, alkaline phosphatase, blood glucose, lactate dehydrogenase, and total cholesterol. The application comprises the following steps: obtaining key features of a chronic kidney disease patient to be predicted; inputting the key features of the chronic kidney disease patient to be predicted into the target machine learning model, and obtaining corresponding vulnerability risk prediction results; the target machine learning model is constructed based on the method for constructing a chronic kidney disease patient vulnerability risk prediction model according to any one of claims 1 to 6.
8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for constructing a chronic kidney disease patient vulnerability risk prediction model according to any one of claims 1-6, or the method for applying the chronic kidney disease patient vulnerability risk prediction model according to claim 7.
9. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the method for constructing a chronic kidney disease patient vulnerability risk prediction model according to any one of claims 1-6, or the method for applying the chronic kidney disease patient vulnerability risk prediction model according to claim 7.
10. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to implement the method for constructing a chronic kidney disease patient vulnerability risk prediction model according to any one of claims 1-6, or the method for applying the chronic kidney disease patient vulnerability risk prediction model according to claim 7.