Gout recurrence prediction method, system, equipment and medium

Abstract

The invention provides a gout recurrence prediction method, system and device and a medium. The method comprises the following steps: acquiring clinical data of gout co-disease patients during hospitalization from an electronic health recording system of a multi-center data set; processing the clinical data according to a standardization method, a minimum-maximum normalization method and a Yeo-Johnson transformation method to obtain candidate clinical features, and screening the extracted candidate clinical features by applying a plurality of feature selectors to screen out clinical features highly related to the gout co-disease recurrence risk; constructing a candidate gout recurrence prediction model for gout recurrence prediction by using a plurality of classifiers; screening the gout recurrence prediction candidate model to obtain a gout recurrence prediction model; acquiring clinical features of the patient; the gout recurrence risk of the patient is predicted according to the clinical features of the patient and the gout recurrence prediction model, the clinical features highly related to the gout co-disease recurrence risk can be accurately screened out, the efficient prediction model is constructed, and therefore reliable auxiliary decision support is provided for clinicians.

Description

Technical Field

[0001] The present application relates to the technical field of gout recurrence prediction, and in particular to a gout recurrence prediction method, system, electronic device and storage medium. Background Art

[0002] Gout comorbidity refers to the state in which gout patients suffer from one or more other diseases at the same time. Gout is an inflammatory arthritis caused by the deposition of urate in and around joints due to increased serum uric acid (SUA) levels. This inflammatory arthritis is characterized by sudden onset of joint pain, swelling and inflammation, and often recurs. Comorbidity (or coexistence of multiple diseases) is a damaged life state in which multiple diseases coexist in the same human body, and it is also a general term for these diseases. Epidemiological studies have shown that the prevalence of comorbidities is high and may continue to rise in the future. For gout patients, they may face a higher risk of comorbidities due to long-term hyperuricemia and gout attacks. These comorbidities may include but are not limited to cardiovascular disease, diabetes, kidney disease, etc., and there is a complex interaction between these diseases and gout.

[0003] The management of patients with gout comorbidities is relatively complex because the treatment and care of multiple diseases need to be considered at the same time. However, there is currently a lack of prediction models for recurrence in patients with gout comorbidities during hospitalization, which increases the difficulty of managing patients with gout comorbidities. Most of the existing prediction models are aimed at outpatients or patients with specific diseases, and there is a lack of effective prediction tools for gout patients with complex conditions and multiple comorbidities during hospitalization. This leads to challenges for doctors in predicting the risk of recurrence in patients with gout comorbidities, making it difficult to develop personalized treatment plans. Summary of the invention

[0004] In view of this, it is necessary to provide a gout recurrence prediction method, system, electronic device and storage medium that can overcome at least one of the above defects.

[0005] In a first aspect, an embodiment of the present application provides a method for predicting gout recurrence, the method comprising:

[0006] Obtain clinical data of gout comorbidity patients during hospitalization from the electronic health record system of the multi-center data set, wherein the clinical data includes basic information, medical history, laboratory test results, diagnosis and treatment process, medication use, length of hospital stay, and follow-up data after discharge of the gout comorbidity patients;

[0007] Processing the clinical data according to standardization, minimum-maximum normalization and Yeo-Johnson transformation methods to obtain candidate clinical features, wherein the candidate clinical features are used to predict the risk of recurrence of gout comorbidities;

[0008] Applying a variety of feature selectors to screen the extracted candidate clinical features to screen out clinical features that are highly correlated with the risk of recurrence of gout comorbidities;

[0009] Apply multiple classifiers to construct candidate gout recurrence prediction models for gout recurrence prediction;

[0010] Screening the candidate models for predicting gout recurrence to obtain a gout recurrence prediction model, wherein the gout recurrence prediction model is a model constructed according to an iterative interpolation method and gradient enhancement based on minimum and maximum scaling and feature selection;

[0011] Obtain the patient's clinical characteristics;

[0012] The patient's gout recurrence risk is predicted based on the patient's clinical characteristics and the gout recurrence prediction model.

[0013] In one embodiment, the gout recurrence prediction method further comprises:

[0014] Obtaining variables in the clinical data;

[0015] Excluding the variables whose missing values ​​are higher than a threshold value to obtain the candidate clinical features;

[0016] Interpolating missing values ​​in the candidate clinical characteristics according to iterative interpolation, mean interpolation, median interpolation and KNN interpolation;

[0017] The candidate clinical features are subjected to collinear transformation according to standardization, minimum-maximum normalization and Yeo-Johnson transformation to improve the stability of the gout recurrence prediction model.

[0018] In one embodiment, screening the candidate models for predicting gout recurrence to obtain a gout recurrence prediction model comprises:

[0019] Get the internal training set and internal validation set;

[0020] Obtaining the area under the receiver operating characteristic curve of the candidate model for predicting gout recurrence according to the internal training set and the internal validation set;

[0021] Performing a preliminary screening of the candidate model for predicting gout recurrence according to the area under the receiver operating characteristic curve to obtain a preliminary screening result;

[0022] Performing a secondary screening on the primary screening results according to the DeLong test to obtain a secondary screening result;

[0023] Obtain external validation sets and prospective validation sets;

[0024] The gout recurrence prediction candidate model in the secondary screening results is verified according to the external validation set and the prospective validation set to obtain the gout recurrence prediction model.

[0025] In one embodiment, the gout recurrence prediction method further comprises:

[0026] Evaluate the linear relationship between the candidate clinical features based on Pearson correlation coefficient and variance inflation factor;

[0027] The candidate clinical features whose variance inflation factor or Pearson correlation coefficient does not meet the requirements are excluded.

[0028] In one embodiment, predicting the patient's gout recurrence risk based on the patient's clinical characteristics and the gout recurrence prediction model further includes:

[0029] Obtaining prediction scores of the gout recurrence prediction model for multiple instances;

[0030] Calculating an optimal classification threshold based on the Youden Index and the prediction score;

[0031] comparing the prediction score to the optimal classification threshold;

[0032] If the prediction score is equal to or greater than the optimal threshold, the instance is classified as positive;

[0033] If the prediction score is less than the optimal threshold, the instance is classified as negative;

[0034] Generate classification decision results based on instance classification.

[0035] In one embodiment, the feature selector includes: conditional information maximization feature extraction, conditional mutual information maximization, dual input symmetric correlation, interaction upper limit, joint mutual information, feature selection based on mutual information, mask image modeling, maximum correlation minimum redundancy, Fisher score, Laplace score, Relief-F algorithm, spectral feature selection, tracking ratio criterion, l2,1-norm minimization based on logical loss, l2,1-norm minimization based on least squares loss, multi-clustering feature selection, robust feature selection, unsupervised discriminant feature selection, correlation-based feature selection, ANOVA F value selection, Gini index, ANOVA T value selection, Alpha investment flow feature selection and minimum absolute shrinkage and selection operator.

[0036] In one embodiment, the classifiers include: logistic regression, K-nearest neighbor classifier, support vector machine, naive Bayes, decision tree, extreme random tree, random forest, adaptive boosting classifier, gradient boosting machine, extreme gradient boosting machine, lightweight gradient boosting machine, classification boosting classifier and artificial neural network.

[0037] In a second aspect, an embodiment of the present application provides a gout recurrence prediction system, which is applied to implement the gout recurrence prediction method as described in the first aspect, and the gout recurrence prediction system includes:

[0038] A clinical data acquisition module is used to obtain clinical data of gout comorbidity patients during hospitalization from an electronic health record system of a multi-center data set, wherein the clinical data includes basic information, medical history, laboratory test results, diagnosis and treatment process, drug use, length of hospitalization, and follow-up data after discharge of the gout comorbidity patients;

[0039] A candidate clinical feature extraction module, used for processing the clinical data according to standardization, minimum maximum normalization and Yeo-Johnson transformation methods to obtain candidate clinical features, wherein the candidate clinical features are used to predict the risk of recurrence of gout comorbidities;

[0040] A clinical feature screening module, used for screening the extracted candidate clinical features according to a plurality of feature selectors, so as to screen out clinical features that are highly correlated with the risk of recurrence of gout comorbidities;

[0041] A candidate model building module, used for building a candidate gout recurrence prediction model based on multiple classifiers;

[0042] A model screening module, used for screening the candidate models for predicting gout recurrence to obtain a gout recurrence prediction model, wherein the gout recurrence prediction model is a model constructed by an iterative interpolation method and gradient enhancement based on minimum and maximum scaling and feature selection;

[0043] A clinical characteristics acquisition module is used to obtain the clinical characteristics of patients;

[0044] The gout recurrence prediction model is used to predict the patient's gout recurrence risk based on the patient's clinical characteristics and the gout recurrence prediction model.

[0045] In a third aspect, an embodiment of the present application provides an electronic device, comprising: a processor; and a memory for storing instructions executable by the processor; wherein the processor is configured to implement the gout recurrence prediction method described in the first aspect when executing the instructions.

[0046] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, comprising instructions, wherein the instructions instruct a device to execute the gout recurrence prediction method as described in the first aspect.

[0047] The gout recurrence prediction method, system, electronic device and storage medium provided in the embodiments of the present application can effectively handle the incomplete and skewed distribution problems in clinical data by comprehensively using clinical data of multi-center data sets and various data processing techniques, such as missing value interpolation, standardization, minimum and maximum normalization, Yeo-Johnson transformation, etc., and improve the quality and stability of the data. By adopting a variety of feature selectors, classifier combinations and collaborative optimization model selection processes, this method can accurately screen out clinical features that are highly correlated with the risk of recurrence of gout comorbidities, and construct an efficient prediction model, thereby providing reliable auxiliary decision support for clinicians. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 A schematic flow chart of a method for predicting gout recurrence provided in one embodiment of the present application.

[0049] Figure 2 It is a schematic diagram of evaluating feature importance using a SHAP value provided in an embodiment of the present application.

[0050] Figure 3 This is a schematic diagram of a gout recurrence prediction system module provided in an embodiment of the present application.

[0051] Figure 4 A schematic diagram of an electronic device provided by an embodiment of the present application.

[0052] Main component symbols

[0053] Gout recurrence prediction system 10

[0054] Clinical Data Acquisition Module 11

[0055] Candidate clinical feature extraction module 12

[0056] Clinical feature screening module 13

[0057] Candidate model building module 14

[0058] Model screening module 15

[0059] Clinical feature acquisition module 16

[0060] Gout recurrence prediction model17

[0061] Electronic devices 20

[0062] Processor 21

[0063] Memory 22

[0064] Method steps S100-S700 DETAILED DESCRIPTION

[0065] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with 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 of the embodiments.

[0066] It should be noted that, in the embodiments of the present application, "at least one" refers to one or more, and "more" refers to two or more. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art in the present application. The terms used in the specification of the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application.

[0067] It should be noted that, in the embodiments of the present application, words such as "first" and "second" are only used for the purpose of distinguishing descriptions, and cannot be understood as indicating or implying relative importance, nor can they be understood as indicating or implying order. Features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a concrete way.

[0068] Based on the implementations in this application, all other implementations obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0069] Gout comorbidity refers to the state in which gout patients suffer from one or more other diseases at the same time. Gout is an inflammatory arthritis caused by the deposition of urate in and around joints due to increased serum uric acid (SUA) levels. This inflammatory arthritis is characterized by sudden onset of joint pain, swelling and inflammation, and often recurs. Comorbidity (or coexistence of multiple diseases) is a damaged life state in which multiple diseases coexist in the same human body, and it is also a general term for these diseases. Epidemiological studies have shown that the prevalence of comorbidities is high and may continue to rise in the future. For gout patients, they may face a higher risk of comorbidities due to long-term hyperuricemia and gout attacks. These comorbidities may include but are not limited to cardiovascular disease, diabetes, kidney disease, etc., and there is a complex interaction between these diseases and gout.

[0070] The management of patients with gout comorbidities is relatively complex because the treatment and care of multiple diseases need to be considered at the same time. However, there is currently no report on the prediction model for recurrence of patients with gout comorbidities during hospitalization, which increases the difficulty of managing patients with gout comorbidities. Most of the existing prediction models are aimed at outpatients or patients with specific diseases, and there is a lack of effective prediction tools for gout patients with complex conditions and multiple comorbidities during hospitalization. This leads to challenges for doctors in predicting the risk of recurrence in patients with gout comorbidities, making it difficult to formulate personalized treatment plans.

[0071] Therefore, the embodiments of the present application provide a gout recurrence prediction method, system, electronic device and storage medium, which can effectively predict the recurrence risk of patients with gout comorbidity during hospitalization. The method comprehensively considers a variety of clinical data such as the patient's basic information, medical history, laboratory test results, and drug use, and uses advanced feature selection and classification algorithms to accurately screen out clinical features that are highly correlated with the risk of recurrence. Through this method, doctors can identify high-risk patients in a timely manner during hospitalization, and take personalized intervention measures to reduce the recurrence rate and improve the patient's treatment effect and prognosis.

[0072] In addition, the gout recurrence prediction system and electronic equipment provided in this application can obtain and process the patient's clinical data in real time in clinical applications, automatically generate prediction results, reduce the workload of medical staff, and improve work efficiency. At the same time, the system supports cross-hospital and multi-center data sharing and verification, ensuring the wide applicability and reliability of the prediction results. Through verification of different hospitals and different populations, the versatility of the model has been further improved, providing a more scientific and accurate treatment plan for patients with gout comorbidities, thereby effectively promoting the development of precision medicine.

[0073] In conjunction with the accompanying drawings, some embodiments of the application are described in detail below. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0074] Figure 1 1 is a flow chart of a method for predicting gout recurrence provided in one embodiment of the present application. Figure 1The gout recurrence prediction method shown includes at least the following steps: S100: Obtain clinical data of gout comorbidity patients during hospitalization from the electronic health record system of a multi-center data set; S200: Process the clinical data according to standardization, minimum and maximum normalization and Yeo-Johnson transformation methods to obtain candidate clinical features; S300: Apply a variety of feature selectors to screen the extracted candidate clinical features to screen out clinical features that are highly correlated with the risk of recurrence of gout comorbidity; S400: Apply a variety of classifiers to construct a candidate model for gout recurrence prediction; S500: Screen the candidate models for gout recurrence prediction to obtain a gout recurrence prediction model; S600: Obtain the patient's clinical characteristics; S700: Predict the patient's gout recurrence risk based on the patient's clinical characteristics and the gout recurrence prediction model.

[0075] S100: Obtain clinical data of patients with gout comorbidities during hospitalization from the electronic health record system of a multicenter dataset.

[0076] In an embodiment of the present application, a gout recurrence prediction method is applied to the prediction of gout recurrence in patients with gout comorbidity. The gout recurrence prediction method includes, in step S100, acquiring clinical data of patients with gout comorbidity during hospitalization from an electronic health record system of a multi-center data set.

[0077] Specifically, clinical data include basic information, medical history, laboratory test results, diagnosis and treatment process, drug use, length of hospital stay and follow-up data after discharge of patients with gout comorbidity. In step S100, hospitalization data of patients with gout comorbidity are first collected from the electronic health record (EHR) system of multiple hospitals. These data are derived from a multi-center data set, including medical records from multiple hospitals, which can provide the model with enough diverse patient samples and increase the applicability and generalization ability of the model. Clinical data include basic information of patients (such as age, gender, etc.), medical history (such as whether they suffer from cardiovascular disease, diabetes and other comorbidities), laboratory test results (such as uric acid levels, renal function, etc.), diagnosis and treatment process (such as hospitalization, surgical records, etc.), drug use (such as gout treatment drugs, other drugs, etc.), length of hospital stay and follow-up data after discharge of patients. These data provide the basis for subsequent feature processing, screening, modeling and other steps.

[0078]

[0079]

[0080] Table 1: Clinical data classification statistics

[0081] Please refer to Table 1, which is a statistical table of clinical data classification provided in the embodiment of the present application. As shown in Table 1, there are 82 variables involved in the present application. It can be understood that the above variables are finally included after excluding variables with missing values ​​exceeding 25%.

[0082] Understandably, the core of this step is to obtain multi-center data, which not only contains basic information about patients, but also covers comprehensive clinical indicators, including various factors that may affect gout recurrence. The use of multi-center data sets can ensure the applicability of the model in different regions and different hospital conditions, thereby improving the accuracy and robustness of the model.

[0083] S200: Process the clinical data according to standardization, min-max normalization and Yeo-Johnson transformation methods to obtain candidate clinical features.

[0084] In an embodiment of the present application, the gout recurrence prediction method includes in step S200 processing the clinical data according to standardization, minimum and maximum normalization and Yeo-Johnson transformation methods to obtain candidate clinical features, wherein the candidate clinical features are used to predict the risk of recurrence of gout comorbidities.

[0085] Specifically, in step S200, preprocessing is performed for different types of clinical data. First, for continuous variables, such as the patient's age, weight, laboratory test results, etc., the data is scaled to the same scale through standardization processing, so that the mean of each variable is 0 and the standard deviation is 1, so as to avoid the excessive influence of certain features on model training due to differences in different dimensions. Secondly, for those features with different distributions, the minimum and maximum normalization method (Min-MaxScaling) is used to convert the data to the [0,1] interval to ensure that the numerical range of each feature is consistent. In addition, for features with skewed distributions (such as blood sugar, uric acid, etc.), the Yeo-Johnson transformation is used to transform the data to make it more consistent with the normal distribution, thereby improving the training effect of the model.

[0086] It is understandable that clinical data preprocessing is the basis for ensuring the accuracy of subsequent analysis. Through these processing methods, the dimension and distribution differences between different features can be eliminated, so that each feature can play a fair role in model training, while also improving the model's adaptability to skewed data and further improving the model's predictive ability.

[0087] S300: Apply multiple feature selectors to screen the extracted candidate clinical features to identify clinical features that are highly correlated with the risk of recurrence of gout comorbidities.

[0088] In an embodiment of the present application, the gout recurrence prediction method includes applying a variety of feature selectors to screen the extracted candidate clinical features in step S300 to screen out clinical features that are highly correlated with the risk of recurrence of gout comorbidities.

[0089] Specifically, in step S300, a variety of feature selectors are applied to screen candidate clinical features. These feature selectors include statistical methods (such as chi-square test, analysis of variance) and model-based methods (such as Lasso regression, decision tree feature importance evaluation, etc.). Through these methods, the correlation between each clinical feature and gout recurrence can be evaluated, and features that contribute greatly to recurrence prediction can be screened out. For example, features with high correlation or that can significantly distinguish between recurrence and non-recurrence will be retained, while redundant or irrelevant features will be eliminated. This can effectively reduce the feature space, reduce computational complexity, and improve the stability and accuracy of the model.

[0090] Please also read Figure 2 , Figure 2 Schematic diagram of SHAP value evaluation of feature importance provided by an embodiment of the present application. Figure 2 As shown, length of stay (LOS) was identified as the most influential factor, followed by glucocorticoids (GCs), uric acid (UA), tophi (Tophus), neutrophils (NEUT), basophils percentage (Basophils%), fibrinogen (Fibrinogen), plateletcrit (Plateletcrit), C-reactive protein (CRP), stroke (Stroke), D-dimer (DD), sodium bicarbonate (NaHCO3), prothrombin time (Prothrombin time), lymphocyte percentage (Lymphocyte percentage), systemic immune-inflammatory index (SII), activated partial thromboplastin time (APTT), pan-immune-inflammatory value (PIV), diuretic (Diuretic), weight change (Weight Changed), and albumin / globulin ratio (AGR). Figure 2 The relationships and interactions between key features are shown.

[0091] It is understandable that feature selection not only helps to reduce overfitting, but also improves the computational efficiency of the model. Through the combination of multiple feature selectors, the most predictive features can be extracted from a large amount of clinical data, making the final constructed model more accurate and reliable in practical applications.

[0092] S400: Apply multiple classifiers to construct a candidate model for predicting gout recurrence.

[0093] In an embodiment of the present application, the gout recurrence prediction method includes applying multiple classifiers to construct a gout recurrence prediction candidate gout recurrence prediction model in step S400.

[0094] Specifically, in step S400, a group of appropriate classifiers, such as Logistic Regression, Support Vector Machine (SVM), Random Forest, Gradient Boosting, etc., will be selected to train the processed and screened clinical features. These classifiers use different algorithms to learn and train data to build multiple candidate models to capture the advantages of different features and classifiers. These models are trained on the same data set to generate multiple recurrence prediction candidate models, providing multiple alternatives for subsequent model screening and optimization.

[0095] It is understandable that the application of multiple classifiers can improve the robustness of the model. Different classifiers may have different advantages and disadvantages when processing data. Through the combination of multiple algorithms, it is possible to mine information in the data from different angles, thereby improving the accuracy of prediction.

[0096] S500: Screening candidate models for predicting gout recurrence to obtain a gout recurrence prediction model.

[0097] In the embodiment of the present application, the gout recurrence prediction method includes screening the gout recurrence prediction candidate model to obtain the gout recurrence prediction model in step S500, and the gout recurrence prediction model is a model constructed by the iterative interpolation method and gradient enhancement based on minimum and maximum scaling and feature selection.

[0098] Specifically, in step S500, multiple candidate models obtained through training are compared in the screening stage, and the optimal recurrence prediction model is finally selected. The selection criteria of the model include: AUC value (area under the curve), accuracy, sensitivity, specificity, F1 score and other evaluation indicators. By combining these indicators, the model with the best performance is screened out. The final recurrence prediction model is constructed based on minimum and maximum scaling, iterative interpolation method of feature selection and gradient enhancement to ensure that the prediction effect of the model has good stability and accuracy on different data sets.

[0099] It is understandable that this screening process is the key to model optimization. By comprehensively evaluating the performance of each candidate model, the model with the best performance in different situations can be selected to ensure high prediction accuracy and stability in practical applications.

[0100] S600: Obtain clinical characteristics of the patient.

[0101] In an embodiment of the present application, the gout recurrence prediction method includes obtaining clinical characteristics of the patient in step S600.

[0102] Specifically, in step S600, clinical characteristics of the patient to be predicted are obtained. These characteristics include the patient's basic information (such as age, gender), medical history (such as whether there is hyperglycemia, cardiovascular disease, etc.), current laboratory test results (such as uric acid level, renal function index, etc.), drug use (such as whether uric acid-lowering drugs are used for a long time, etc.), and treatment plans and clinical processes during hospitalization. Through these characteristics, the necessary input information can be provided for the prediction model.

[0103] It is understandable that obtaining the patient's clinical characteristics is a prerequisite for predictive analysis, ensuring that the ventilation recurrence prediction method can accurately predict recurrence based on real-time patient information. These characteristics are the input of the prediction model and affect the accuracy of the model output.

[0104] S700: Predict the patient's risk of gout recurrence based on their clinical characteristics and gout recurrence prediction model.

[0105] In an embodiment of the present application, the gout recurrence prediction method includes, in step S700, predicting the patient's gout recurrence risk based on the patient's clinical characteristics and a gout recurrence prediction model.

[0106] Specifically, in step S700, the clinical characteristics of the patient obtained in the previous step are used and input into the gout recurrence prediction model constructed previously. The model predicts based on the input characteristics and obtains a probability score for recurrence. Based on the probability output by the model, it can be determined whether the patient is at risk of recurrence, and decision support can be provided to the doctor to help him formulate a more appropriate treatment plan.

[0107] It can be understood that this step provides a quantitative prediction of the risk of recurrence by combining the patient's clinical characteristics with the prediction model. Doctors can judge the trend of the patient's condition based on the prediction results and make timely interventions, thereby improving the patient's prognosis.

[0108] In an embodiment of the present application, the gout recurrence prediction method also includes obtaining variables in clinical data. Exclude variables with missing values ​​higher than a threshold value to obtain candidate clinical features. Interpolate the missing values ​​in the candidate clinical features according to iterative interpolation, mean interpolation, median interpolation, and KNN interpolation. Perform collinear transformation on the candidate clinical features according to standardization, minimum and maximum normalization, and Yeo-Johnson transformation to improve the stability of the gout recurrence prediction model.

[0109] Specifically, in the present application embodiment, the implementation of the gout recurrence prediction method also includes extracting multiple variables from clinical data, and these variables include the patient's basic information, clinical symptoms, medical history, drug use, laboratory test results, etc. In order to ensure the integrity and effectiveness of the data, first the missing values ​​will be processed, excluding the variables whose missing value ratio exceeds the set threshold. Subsequently, interpolation methods (such as iterative interpolation, mean interpolation, median interpolation and KNN interpolation) are used to fill these missing values ​​to ensure that each feature variable can provide enough information for subsequent analysis. For the processed candidate features, collinearity conversion will also be performed to reduce the multicollinearity problem between different variables, so as to improve the stability and accuracy of model prediction. Standardization, minimum and maximum normalization and Yeo-Johnson transformation methods are used to balance the scale difference between variables and process the skewed distribution of data, to ensure the effectiveness of model training.

[0110]

[0111]

[0112]

[0113] Table 2: Variable linearization results

[0114] Please refer to Table 2, which is a variable linearization representation result provided in an embodiment of the present application, wherein O represents the feature retained after solving the collinearity problem.

[0115] It can be understood that by using the above interpolation method and data conversion means, a candidate feature set with high quality and stability can be generated. These features are used in subsequent model training to ensure that the constructed prediction model can accurately reflect the risk of recurrence in patients with gout comorbidities and can effectively deal with missing data and collinearity problems.

[0116] In an embodiment of the present application, screening a candidate model for predicting gout recurrence to obtain a gout recurrence prediction model includes obtaining an internal training set and an internal validation set. The area under the receiver operating characteristic curve of the candidate model for predicting gout recurrence is obtained according to the internal training set and the internal validation set. The candidate model for predicting gout recurrence is preliminarily screened according to the area under the receiver operating characteristic curve to obtain a preliminarily screening result. The preliminarily screening result is secondary screened according to the DeLong test to obtain a secondary screening result. An external validation set and a prospective validation set are obtained. The candidate model for predicting gout recurrence in the secondary screening result is verified according to the external validation set and the prospective validation set to obtain a gout recurrence prediction model.

[0117] Specifically, in an embodiment of the present application, the gout recurrence prediction method further relates to screening out the best prediction model. First, the most suitable model is selected from multiple candidate models. This process includes extracting the area under the receiver operating characteristic curve (ROC) (AUC) from the internal training set and the internal validation set, evaluating the classification effect of the model by AUC, and performing preliminary screening. Then, the screening results are statistically tested using the DeLong test to further confirm the validity of the model. On this basis, an external validation set and a prospective validation set are also obtained to verify the candidate model after secondary screening to finally determine the best prediction model.

[0118] It is understandable that the core purpose of this step is to ensure that the selected prediction model has wide applicability and high generalization ability through evaluation of different datasets (including internal and external validation sets), thereby improving the accuracy of gout recurrence prediction.

[0119] In an embodiment of the present application, the gout recurrence prediction method further includes evaluating the linear relationship between candidate clinical features according to the Pearson correlation coefficient and the variance inflation factor, and excluding candidate clinical features whose variance inflation factor or Pearson correlation coefficient does not meet the requirements.

[0120] Specifically, the present application also involves optimizing model training by calculating the correlation between candidate clinical features. The linear relationship between candidate features is evaluated using the Pearson correlation coefficient (PCC) and variance inflation factor (VIF), and features that do not meet the requirements are excluded. If the correlation coefficient of some candidate features is too high (such as PCC>0.8) or the VIF value is too large (such as VIF>10), it means that these features have multicollinearity, which may cause model instability, so these features need to be excluded from the analysis.

[0121] It can be understood that through this feature screening method, the negative impact of multicollinearity on model training can be eliminated, ensuring that the selected features have stronger independence and predictive ability, thereby improving the prediction effect and stability of the model.

[0122] In one embodiment, the gout recurrence risk of a patient is predicted based on the patient's clinical characteristics and a gout recurrence prediction model, and the gout recurrence prediction model is also used to obtain the prediction scores of multiple instances. The optimal classification threshold is calculated based on the Youden index and the prediction score. The prediction score is compared with the optimal classification threshold. If the prediction score is equal to or greater than the optimal threshold, the instance is classified as positive. If the prediction score is less than the optimal threshold, the instance is classified as negative. A classification decision result is generated based on the instance classification.

[0123] Specifically, in some embodiments, based on the patient's clinical characteristics and gout recurrence prediction model, the system will calculate the prediction scores of multiple instances. The optimal classification threshold is calculated based on Youden's index, and the prediction score is compared with the optimal threshold. If the prediction score is greater than or equal to the optimal threshold, the system classifies the instance as positive; otherwise, it is classified as negative. In this way, the system can generate specific classification decision results.

[0124] It can be understood that by comparing the prediction score with the optimal threshold, it can be ensured that the classification decision maximizes the balance between the true positive rate and the true negative rate, thereby improving the clinical application value of the model and enabling doctors to more accurately determine the recurrence risk of patients with gout comorbidities and formulate personalized treatment plans accordingly.

[0125] In one embodiment, the feature selector includes: conditional information maximization feature extraction, conditional mutual information maximization, dual input symmetric correlation, interaction upper limit, joint mutual information, feature selection based on mutual information, mask image modeling, maximum correlation minimum redundancy, Fisher score, Laplace score, Relief-F algorithm, spectral feature selection, tracking ratio criterion, l2,1-norm minimization based on logical loss, l2,1-norm minimization based on least squares loss, multi-clustering feature selection, robust feature selection, unsupervised discriminant feature selection, correlation-based feature selection, ANOVAF value selection, Gini index, ANOVAT value selection, Alpha investment flow feature selection and minimum absolute shrinkage and selection operator.

[0126] Specifically, in the embodiment of the present application, the feature selector includes a variety of information theory methods, such as conditional information maximization, conditional mutual information maximization, maximum correlation minimum redundancy, etc., which can identify the most predictive features. In addition, methods based on statistical analysis (such as Fisher score, ANOVAF value selection, etc.) and machine learning (such as Relief-F algorithm, l2,1-norm minimization based on least squares loss, etc.) are also used to enhance the effect of feature selection.

[0127] It can be understood that through the combined use of multiple feature selection methods, features that are highly correlated with the risk of recurrence of gout comorbidities can be effectively extracted, the predictive performance and explanatory ability of the model can be improved, and the final model can be ensured to accurately reflect the impact of clinical characteristics on the risk of recurrence.

[0128] In one embodiment, the classifiers include: logistic regression, K-nearest neighbor classifier, support vector machine, naive Bayes, decision tree, extreme random tree, random forest, adaptive boosting classifier, gradient boosting machine, extreme gradient boosting machine, lightweight gradient boosting machine, classification boosting classifier and artificial neural network.

[0129] Specifically, in the embodiments of the present application, the classifier includes common machine learning algorithms, such as logistic regression, K-nearest neighbor classifier, support vector machine (SVM), naive Bayes, decision tree, random forest, etc. In addition, a variety of reinforcement learning methods (such as gradient boosting machine, extreme gradient boosting machine, lightweight gradient boosting machine, etc.) and artificial neural networks are also used to improve the prediction accuracy of the model.

[0130] It can be understood that by using different types of classifiers, the recurrence risk of patients can be evaluated from multiple perspectives, further improving the robustness and generalization ability of the model, thereby ensuring the effectiveness of the model in clinical applications.

[0131] The following is an exemplary embodiment to describe the process of constructing a gout recurrence prediction model in the gout recurrence prediction method provided in the present application.

[0132] This embodiment aims to build an accurate gout recurrence prediction model by integrating multiple data processing methods (interpolation, feature conversion, feature selection and classifier). The model uses a multi-center data set, which comes from Nanfang Hospital, Ganzhou People's Hospital, Southern Medical University Hospital of Integrated Traditional Chinese and Western Medicine, Taishan People's Hospital and Dongguan Traditional Chinese Medicine Hospital. Through feature selection and classifier combined training, a recurrence prediction model with high generalization ability is finally constructed.

[0133] First, clinical data of gout comorbidity patients during hospitalization were obtained from the electronic health record system, including basic patient information, medical history, laboratory test results, medication use, length of hospital stay, and follow-up data. Continuous variables were standardized, minimum and maximum normalized, and Yeo-Johnson transformed to handle skewed distribution variables to ensure data quality.

[0134] Subsequently, 24 feature selectors were used to screen candidate clinical features. After all features were screened, features with strong collinearity (PCC>0.8 or VIF>10) were removed to reduce the impact of multicollinearity on the model.

[0135] Next, four imputation methods (Iterative Imputer, Mean Imputer, Median Imputer, and KNN Imputer) were used to impute missing values. The data was further processed by combining three feature transformation methods (Min-Max Scaling, Standardization, and Yeo-Johnson Transformation). 3744 models were constructed using a combination of 13 classifiers (such as Logistic Regression, KNN, SVM, etc.) and 24 feature selectors.

[0136] Next, after screening the internal validation set, overfitting models (models with AUC differences greater than 10%) were removed, and 1,664 models were screened out. Based on the AUC value and DeLong test results, the top 20 models were further screened out, and then 5 models were screened out for the final screening. Finally, the IterImp_MM_FS_GB model performed best on all data sets, especially on external data sets and prospective data sets, with AUC values ​​exceeding 0.74, and high accuracy and stability in predicting recurrence. Therefore, the model constructed with the iterative interpolation method and gradient enhancement based on minimum and maximum scaling and feature selection was selected as the best model.

[0137]

[0138]

[0139]

[0140] Table 3: Parameters of the top five models

[0141] Please also refer to Table 3, which is a table of parameters of the top five models. As shown in Table 3, the model built by iterative interpolation method with gradient boosting based on min-max scaling and feature selection performs best.

[0142] Finally, on the training dataset, the model constructed by iterative interpolation method with minimum and maximum scaling and feature selection and gradient enhancement had an AUC of 0.832, an accuracy of 0.745, an NPV of 0.876, and an F1 score of 0.630; on the external validation set, the AUC was 0.742, the accuracy was 0.681, the NPV was 0.867, and the F1 score was 0.588; on the prospective dataset, the model had an AUC of 0.744, an accuracy of 0.749, an NPV of 0.917, and an F1 score of 0.383. The model combining iterative interpolation with gradient enhancement showed excellent discrimination ability and high stability.

[0143] Figure 3 This is a gout recurrence prediction system provided by an embodiment of the present application. Figure 3 The gout recurrence prediction system 10 shown includes at least the following parts: a clinical data acquisition module 11, a candidate clinical feature extraction module 12, a clinical feature screening module 13, a candidate model construction module 14, a model screening module 15, a clinical feature acquisition module 16 and a gout recurrence prediction model 17.

[0144] In the embodiment of the present application, the clinical data acquisition module 11 is used to obtain clinical data of gout comorbidity patients during hospitalization from the electronic health record system of the multi-center data set. The clinical data includes the basic information, medical history, laboratory test results, diagnosis and treatment process, drug use, length of hospitalization and follow-up data after discharge of the gout comorbidity patients. For specific acquisition methods, please refer to Figure 1 And the corresponding description thereof, this application will not repeat them here.

[0145] In the embodiment of the present application, the candidate clinical feature extraction module 12 is used to process the clinical data according to the standardization, minimum and maximum normalization and Yeo-Johnson transformation method to obtain candidate clinical features, which are used to predict the recurrence risk of gout comorbidity. For specific acquisition methods, please refer to Figure 1 And the corresponding description thereof, this application will not repeat them here.

[0146] In the embodiment of the present application, the clinical feature screening module 13 is used to screen the extracted candidate clinical features according to a plurality of feature selectors to screen out clinical features that are highly correlated with the risk of recurrence of gout comorbidities. Figure 1 And the corresponding description thereof, this application will not repeat them here.

[0147] In the embodiment of the present application, the candidate model construction module 14 is used to construct a candidate gout recurrence prediction model based on multiple classifiers. Figure 1 And the corresponding description thereof, this application will not repeat them here.

[0148] In the embodiment of the present application, the model screening module 15 is used to screen the candidate models for predicting gout recurrence to obtain a gout recurrence prediction model, which is a model constructed by an iterative interpolation method and gradient enhancement based on minimum and maximum scaling and feature selection. For specific screening methods, please refer to Figure 1 And the corresponding description thereof, this application will not repeat them here.

[0149] In the embodiment of the present application, the clinical characteristics acquisition module 16 is used to obtain the clinical characteristics of the patient. Figure 1 And the corresponding description thereof, this application will not repeat them here.

[0150] In the embodiment of the present application, the gout recurrence prediction model 17 is used to predict the gout recurrence risk of the patient based on the clinical characteristics of the patient and the gout recurrence prediction model. Figure 1 And the corresponding description thereof, this application will not repeat them here.

[0151] The gout recurrence prediction system 10 provided in the embodiment of the present application ensures the good generalization ability of the prediction model on different data sets through multiple rounds of verification and tuning of the model, including the use of internal training sets, validation sets, external validation sets and prospective validation sets. By optimizing the classification threshold in combination with the Youden index, it is possible to accurately distinguish the high and low risks of recurrence, provide more accurate recurrence predictions, and help doctors formulate personalized treatment plans. Ultimately, this system can provide a scientific, accurate, and clinically operable solution for predicting the recurrence risk of patients with gout comorbidities, significantly improving clinical management and patient prognosis.

[0152] Figure 4 This is an electronic device 20 provided by an embodiment of the present application. Figure 4 As shown, the electronic device 20 includes at least the following parts: a processor 21 and a memory 22.

[0153] In the embodiment of the present application, the memory 22 is used to store instructions executable by the processor 21, and the processor 21 is configured to implement the following when executing the instructions: Figure 1 The method shown is a method for predicting gout recurrence.

[0154] In an embodiment of the present application, a computer-readable storage medium includes instructions, and the instructions instruct a device to execute the method of the first aspect. For example, the instructions instruct the device to execute Figure 1 The gout recurrence prediction method shown in steps S100 to S700.

[0155] The program working in the electronic device 20 involved in one embodiment of the present application can be a program (a program that enables a computer to function) that controls a central processing unit (CPU) and the like to realize the functions of the above-mentioned implementation method involved in one scheme of the present invention. Then, the information processed by these devices is temporarily stored in a random access memory (RAM) when it is processed, and then stored in various ROMs such as a read-only memory (Flash ROM) and a hard disk drive (HDD), and is read, corrected, and written by the CPU as needed.

[0156] It should be noted that a part of the electronic device 20 of the above embodiment may be implemented by a computer. In this case, a program for implementing the control function may be recorded in a computer-readable recording medium, and the program recorded in the recording medium may be read into a computer system and executed.

[0157] It should be noted that the "computer system" mentioned here refers to a computer system built into the electronic device 20, and adopts a computer system including hardware such as OS and peripheral devices. In addition, "computer-readable recording medium" refers to removable media such as floppy disks, magneto-optical disks, ROMs, CD-ROMs, and storage devices such as hard disks built into the computer system.

[0158] Furthermore, "computer-readable recording media" may include: media that dynamically store programs for a short period of time, such as communication lines when sending programs via networks such as the Internet or communication lines such as telephone lines; and media that store programs for a fixed period of time, such as volatile memories inside computer systems that serve as servers or clients in this case. In addition, the above-mentioned program may be a program for realizing a part of the above-mentioned functions, or a program that can realize the above-mentioned functions by combining with a program already recorded in a computer system.

[0159] In addition, the electronic device 20 in the above embodiment can also be implemented as a collection (device group) composed of multiple devices. Each device constituting the device group can have a part or all of the functions or functional blocks of the electronic device 20 in the above embodiment. As a device group, it is sufficient to have all the functions or functional blocks of the electronic device 20.

[0160] It is understandable that the gout recurrence prediction method, system 10, electronic device 20 and storage medium provided in the embodiment of the present application can effectively handle the incomplete and skewed distribution problems in clinical data by comprehensively using the clinical data of multi-center data sets and various data processing techniques, such as missing value interpolation, standardization, minimum and maximum normalization, Yeo-Johnson transformation, etc., and improve the quality and stability of the data. By adopting a variety of feature selectors, classifier combinations and collaborative optimization model selection processes, the method can accurately screen out clinical features that are highly correlated with the risk of recurrence of gout comorbidities, and construct an efficient prediction model, thereby providing reliable auxiliary decision support for clinicians.

[0161] In addition, the embodiment of the present application also ensures the good generalization ability of the prediction model on different data sets through multiple rounds of verification and tuning of the model, including the use of internal training sets, validation sets, external validation sets, and prospective validation sets. By combining the Youden index to optimize the classification threshold, this method can accurately distinguish the high and low risks of recurrence, provide more accurate recurrence predictions, and help doctors develop personalized treatment plans. Ultimately, this system can provide a scientific, accurate, and clinically feasible solution for predicting the recurrence risk of patients with gout comorbidities, significantly improving clinical management and patient prognosis.

[0162] Those skilled in the art should recognize that the above embodiments are only used to illustrate the present application and are not intended to be limiting of the present application. As long as they are within the spirit and scope of the present application, appropriate changes and modifications to the above embodiments are within the scope of protection claimed in the present application.

Claims

1. A gout recurrence prediction method, applied to gout recurrence prediction in patients with gout comorbidities, characterized in that: The gout recurrence prediction method comprises: Obtain clinical data of gout comorbidity patients during hospitalization from the electronic health record system of the multi-center data set, wherein the clinical data includes basic information, medical history, laboratory test results, diagnosis and treatment process, medication use, length of hospital stay, and follow-up data after discharge of the gout comorbidity patients; Processing the clinical data according to standardization, minimum-maximum normalization and YeoJohnson transformation methods to obtain candidate clinical features, wherein the candidate clinical features are used to predict the risk of recurrence of gout comorbidities; Applying a variety of feature selectors to screen the extracted candidate clinical features to screen out clinical features that are highly correlated with the risk of recurrence of gout comorbidities; Apply multiple classifiers to construct candidate gout recurrence prediction models for gout recurrence prediction; Screening the candidate models for predicting gout recurrence to obtain a gout recurrence prediction model, wherein the gout recurrence prediction model is a model constructed by an iterative interpolation method and gradient enhancement based on minimum and maximum scaling and feature selection; Obtain the patient's clinical characteristics; The patient's gout recurrence risk is predicted based on the patient's clinical characteristics and the gout recurrence prediction model.

2. The method for predicting gout recurrence according to claim 1, characterized in that: The gout recurrence prediction method also includes: Obtaining variables in the clinical data; Excluding the variables whose missing values ​​are higher than a threshold value to obtain the candidate clinical features; Interpolating missing values ​​in the candidate clinical characteristics according to iterative interpolation, mean interpolation, median interpolation and K NN interpolation; The candidate clinical features are subjected to collinear transformation according to standardization, minimum-maximum normalization and YeoJohnson transformation to improve the stability of the gout recurrence prediction model.

3. The method for predicting gout recurrence according to claim 2, characterized in that: The step of screening the candidate model for predicting gout recurrence to obtain a gout recurrence prediction model comprises: Get the internal training set and internal validation set; Obtaining the area under the receiver operating characteristic curve of the candidate model for predicting gout recurrence according to the internal training set and the internal validation set; Performing a preliminary screening of the candidate model for predicting gout recurrence according to the area under the receiver operating characteristic curve to obtain a preliminary screening result; Performing a secondary screening on the primary screening results according to the De Long test to obtain a secondary screening result; Obtain external validation sets and prospective validation sets; The gout recurrence prediction candidate model in the secondary screening results is verified according to the external validation set and the prospective validation set to obtain the gout recurrence prediction model.

4. The method for predicting gout recurrence according to claim 2, characterized in that: The gout recurrence prediction method also includes: Evaluate the linear relationship between the candidate clinical features according to the Pearson correlation coefficient and the variance inflation factor; The candidate clinical features whose variance inflation factor or Pearson correlation coefficient does not meet the requirements are excluded.

5. The method for predicting gout recurrence according to claim 1, characterized in that: The method of predicting the patient's gout recurrence risk based on the patient's clinical characteristics and the gout recurrence prediction model also includes: Obtaining prediction scores of the gout recurrence prediction model for multiple instances; Calculating an optimal classification threshold based on the Youden Index and the prediction score; comparing the prediction score to the optimal classification threshold; If the prediction score is equal to or greater than the optimal threshold, the instance is classified as positive; If the prediction score is less than the optimal threshold, the instance is classified as negative; Generate classification decision results based on instance classification.

6. The method for predicting gout recurrence according to any one of claims 1 to 5, characterized in that: The feature selector includes: conditional information maximization feature extraction, conditional mutual information maximization, dual input symmetric correlation, interaction upper limit, joint mutual information, feature selection based on mutual information, mask image modeling, maximum correlation minimum redundancy, Fisher score, Laplace score, ReliefF algorithm, spectral feature selection, tracking ratio criterion, l 2,1-norm minimization based on logical loss, l 2,1-norm minimization based on least squares loss, multi-clustering feature selection, robust feature selection, unsupervised discriminant feature selection, feature selection based on correlation, ANOVAF value selection, Gini index, ANOVAT value selection, Alpha investment flow feature selection and minimum absolute shrinkage and selection operator.

7. The method for predicting gout recurrence according to any one of claims 1 to 5, characterized in that: The classifiers include: logistic regression, K-nearest neighbor classifier, support vector machine, naive Bayes, decision tree, extreme random tree, random forest, adaptive boosting classifier, gradient boosting machine, extreme gradient boosting machine, lightweight gradient boosting machine, classification boosting classifier and artificial neural network.

8. A gout recurrence prediction system, used to implement the gout recurrence prediction method according to any one of claims 1 to 7, characterized in that: The gout recurrence prediction system comprises: A clinical data acquisition module is used to obtain clinical data of gout comorbidity patients during hospitalization from an electronic health record system of a multi-center data set, wherein the clinical data includes basic information, medical history, laboratory test results, diagnosis and treatment process, drug use, length of hospitalization, and follow-up data after discharge of the gout comorbidity patients; A candidate clinical feature extraction module, used for processing the clinical data according to standardization, minimum and maximum normalization and YeoJohnson transformation methods to obtain candidate clinical features, wherein the candidate clinical features are used to predict the risk of recurrence of gout comorbidities; A clinical feature screening module, used for screening the extracted candidate clinical features according to a plurality of feature selectors, so as to screen out clinical features that are highly correlated with the risk of recurrence of gout comorbidities; A candidate model building module, used for building a candidate gout recurrence prediction model based on multiple classifiers; A model screening module, used for screening the candidate models for predicting gout recurrence to obtain a gout recurrence prediction model, wherein the gout recurrence prediction model is a model constructed by an iterative interpolation method and gradient enhancement based on minimum and maximum scaling and feature selection; A clinical characteristics acquisition module is used to obtain the clinical characteristics of patients; The gout recurrence prediction model is used to predict the patient's gout recurrence risk based on the patient's clinical characteristics and the gout recurrence prediction model.

9. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; Wherein, the processor is configured to implement the gout recurrence prediction method described in any one of claims 1 to 7 when executing the instructions.

10. A computer-readable storage medium, characterized in that: Comprising instructions, wherein the instructions instruct the device to execute the gout recurrence prediction method as described in any one of claims 1 to 7.

Images