Renal dialysis prediction method, device, equipment and storage medium

Through multi-classifier model screening and training based on historical medical record information, the risk of postoperative renal dialysis is accurately predicted, which solves the problem of difficult to predict postoperative renal dialysis risks in the prior art, and improves the efficiency of patient prognosis and medical management.

CN117912700BActive Publication Date: 2024-11-15THE FIRST AFFILIATED HOSPITAL OF JINAN UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410150751.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-02-02
Publication Date
2024-11-15
Estimated Expiration
2044-02-02

AI Technical Summary

Technical Problem

The prior art is difficult to accurately predict the risk of postoperative renal dialysis, resulting in poor patient prognosis and increased medical expenses.

Method used

Develop a kidney dialysis prediction method, by obtaining preoperative and intraoperative clinical feature information of patients, and establishing it using multiple preset classifiers (such as LightGBM, XGBoost, SVM, RF, AdaBoost, and Logistic regression) based on historical medical record information. Target prediction model, screen relevant clinical characteristics and make predictions.

Benefits of technology

The rapid and accurate prediction of postoperative renal dialysis risks is achieved, reducing the impact of physician experience, improving the stability and accuracy of predictions, and supporting early intervention and postoperative safety management.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117912700B_ABST
    Figure CN117912700B_ABST
Patent Text Reader

Abstract

The present invention provides a method, device, equipment and storage medium for predicting renal dialysis. The method comprises: S101: obtaining clinical information of a predicted object, wherein clinical characteristics corresponding to the clinical information include preoperative characteristics and intraoperative characteristics; S102: inputting the clinical information into a target prediction model, and obtaining the renal dialysis risk of the predicted object according to the output information of the target prediction model, wherein the target prediction model is obtained by training multiple preset classifiers based on historical medical record information. The present invention can quickly obtain the renal dialysis risk of the predicted object according to the clinical information of the predicted object immediately after the operation is completed, so as to facilitate early intervention and protect the postoperative safety of the predicted object. Moreover, the prediction process does not need to rely on the doctor's experience, and effectively reduces the influence of the doctor's subjective experience, and has higher stability and accuracy.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of disease prediction, and in particular to a method, device, equipment and storage medium for predicting renal dialysis. Background Art

[0002] Acute kidney injury is a common complication after cardiovascular surgery. It is a syndrome characterized by a sharp decline in renal function. It has a high incidence rate and is associated with increased mortality and medical expenses, especially acute kidney injury that requires renal dialysis after surgery. Renal dialysis is a more serious adverse event than renal insufficiency, with a short-term mortality rate of up to 64%. It is an important cause of death and poor prognosis for patients in the intensive care unit. Given the huge economic and psychological burden and prognostic impact of postoperative renal dialysis. However, the risk at the patient level is currently difficult to predict, and there is an urgent need to develop and validate an accurate and clinically meaningful renal dialysis risk prediction model to ultimately improve patient prognosis. Summary of the invention

[0003] In order to overcome the deficiencies of the prior art, the present invention proposes a method, device, equipment and storage medium for predicting renal dialysis, which solves the problem that the current renal dialysis risk is difficult to predict and the patient's prognosis is poor.

[0004] To solve the above problems, a technical solution adopted by the present invention is: a renal dialysis prediction method, the renal dialysis prediction method comprising: S101: obtaining clinical information of the predicted object, the clinical information comprising information corresponding to clinical characteristics, the clinical characteristics comprising preoperative characteristics and intraoperative characteristics; S102: inputting the clinical information into a target prediction model, and obtaining the renal dialysis risk of the predicted object according to the output information of the target prediction model, the target prediction model being obtained by training multiple preset classifiers based on historical medical record information.

[0005] Furthermore, multiple preset classifiers are trained based on historical medical record information, including: obtaining historical clinical feature information based on the historical medical record information of historical patients; training multiple preset classifiers based on the historical clinical feature information to obtain a prediction model, and determining a target prediction model based on the performance parameters of each prediction model, the preset classifiers include at least two of LightGBM, XGBoost, SVM, RF, AdaBoost and Logistic regression.

[0006] Furthermore, the method of obtaining clinical characteristic information based on historical medical records of historical patients includes: processing the historical medical records information based on a first preset standard to generate training data, wherein the first preset standard includes at least one of treatment time, disease type, degree of completeness of clinical data, and postoperative status; obtaining predictive variables corresponding to the training data, and filtering clinical characteristics from the predictive variables according to preset filtering conditions, wherein the preset filtering conditions include at least one of high consistency, non-redundancy, and correlation.

[0007] Furthermore, the method of screening clinical characteristics from the predictor variables according to preset screening conditions includes: eliminating predictor variables with a variance inflation factor greater than a first preset value and a correlation coefficient greater than a second preset value to obtain a second variable; obtaining clinical characteristics from the second variable according to a preset method, the preset method including a LASSO regression method, the preoperative characteristics in the clinical characteristics including at least one of serum creatinine, platelets, history of cardiovascular surgery, right coronary artery involvement, and terminal site of dissection involvement, and the intraoperative characteristics including at least one of urine volume and estimated blood loss.

[0008] Furthermore, the method of training a preset classifier according to the historical clinical feature information to obtain a prediction model includes: obtaining historical clinical feature information from the historical case information according to the clinical features, and forming a training cohort and a test cohort based on the historical clinical feature information; and training and verifying the preset classifier according to the training cohort to obtain a prediction model.

[0009] Furthermore, the determining of the target prediction model based on the performance parameters of each prediction model includes: obtaining the performance parameters of each test cohort based on the test cohort and the training cohort, the performance parameters including at least one of discrimination, calibration, and clinical utility evaluation.

[0010] Furthermore, obtaining the renal dialysis risk of the predicted object according to the output information of the target prediction model includes: displaying the probability of the predicted object developing renal dialysis according to the output information, and displaying the correlation between different clinical characteristics and the probability according to the output information.

[0011] Based on the same inventive concept, the present invention also proposes a renal dialysis prediction device, comprising: a clinical information acquisition module, used to obtain clinical information of the predicted object, the clinical information includes information corresponding to clinical characteristics, and the clinical characteristics include preoperative characteristics and intraoperative characteristics; a prediction module, used to input the clinical characteristic information into a target prediction model, and obtain the renal dialysis risk of the predicted object according to the output information of the target prediction model, and the target prediction model is obtained by training a preset classifier based on historical medical record information.

[0012] Based on the same inventive concept, the present invention also proposes an electronic device, which includes a processor and a memory, wherein the memory stores a computer program, and the computer program is used by the processor to implement the renal dialysis prediction method as described above.

[0013] Based on the same inventive concept, the present invention further proposes a computer-readable storage medium, wherein the computer-readable storage medium stores program data, and the program data is used to execute the renal dialysis prediction method as described above.

[0014] The beneficial effect of the present invention is that it provides a method for predicting renal dialysis, obtains clinical information related to the predicted object and the preoperative characteristics and intraoperative characteristics, inputs the clinical information into a target prediction model obtained by training multiple preset classifiers with historical case information, and determines the renal dialysis risk of the predicted object through the output information of the target prediction model. The present application can quickly obtain the renal dialysis risk of the predicted object based on the clinical information of the predicted object immediately after the operation is completed, which is convenient for early intervention and protects the postoperative safety of the predicted object. In addition, the prediction process does not need to rely on the doctor's experience, effectively reduces the influence of the doctor's experience, and has higher stability and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 A flowchart of an embodiment of a method for predicting renal dialysis according to the present invention;

[0016] Figure 2 It is a structural diagram of an embodiment of a renal dialysis prediction device of the present invention;

[0017] Figure 3 It is a structural diagram of an embodiment of an electronic device of the present invention. DETAILED DESCRIPTION

[0018] The following describes the implementation methods of the present application through specific specific examples, and those skilled in the art can easily understand other advantages and effects of the present application from the contents disclosed in this specification. The present application can also be implemented or applied through other different specific implementation methods, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present application. It should be noted that the various embodiments of the present disclosure generally described and shown in the accompanying drawings here can be combined with each other without conflict, and the structural components or functional modules therein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present disclosure provided in the accompanying drawings is not intended to limit the scope of the present disclosure claimed for protection, but merely represents the selected embodiments of the present disclosure. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present disclosure.

[0019] The terms used in the disclosure of this application are only for the purpose of describing specific embodiments and are not intended to limit the present disclosure. The singular forms of "a", "said" and "the" used in the present disclosure and the appended claims are also intended to include plural forms unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0020] Acute type A aortic dissection is a life-threatening disease with a 30-day mortality rate of 10-35% in patients undergoing surgery and a high incidence of major postoperative complications. Patients undergoing surgery for acute type A aortic dissection may experience a common complication called postoperative acute kidney injury, which is caused by poor renal perfusion, complexity of the surgery, and significant trauma. Patients who develop postoperative acute kidney injury have high mortality, prolonged hospital stay, high medical costs, and poor quality of life, especially those who develop acute kidney injury requiring dialysis. Renal dialysis is a more serious adverse event than renal insufficiency, with a short-term mortality rate of up to 64%. Early diagnosis and perioperative risk management can help reduce mortality. Given the huge burden and prognostic significance of postoperative dialysis, there is an urgent need to be able to quickly and accurately predict patients' risk of renal dialysis to improve patient outcomes.

[0021] See also Figure 1 ,in, Figure 1 FIG. 1 is a flow chart of an embodiment of a method for predicting renal dialysis according to the present invention. Figure 1 The renal dialysis prediction method of the present invention is described in detail.

[0022] Optionally, the renal dialysis prediction method is implemented based on a target prediction model, wherein the target prediction model is obtained by training multiple preset classifiers based on historical case information.

[0023] Optionally, multiple preset classifiers are trained based on historical medical record information, including: obtaining historical clinical feature information based on historical medical record information of historical patients; training multiple preset classifiers based on historical clinical feature information to obtain prediction models, and determining a target prediction model based on performance parameters of each prediction model.

[0024] Specifically, the preset classifiers include at least two of LightGBM, XGBoost, SVM, RF, AdaBoost and Logistic regression, multiple prediction models are obtained through the above preset classifiers, and a target prediction model is screened from the multiple prediction models.

[0025] Optionally, clinical characteristic information is obtained based on historical medical records of historical patients, including: generating training data by processing historical medical records based on a first preset standard, the first preset standard including at least one of treatment time, disease type, degree of completeness of clinical data, and postoperative status; obtaining predictive variables corresponding to the training data, and screening clinical characteristics from the predictive variables according to preset screening conditions, the preset screening conditions including at least one of high consistency, non-redundancy, and correlation.

[0026] Optionally, information that does not meet the requirements in historical case information is eliminated according to a first preset standard to obtain medical records that can be used for training, and data related to clinical characteristics are obtained from the medical records and used as training data.

[0027] In one embodiment, the historical medical record information is the electronic medical record data of 613 patients with acute type A aortic dissection who underwent surgical treatment between October 8, 2016 and June 30, 2021. These patients were diagnosed with acute type A aortic dissection and the affected blood vessels were determined using computed tomography angiography (CTA). After obtaining the electronic medical record data, the medical record data of some patients are excluded according to the first preset standard. Specifically, the first preset standard includes: (1) patients who have received dialysis treatment within 90 days before surgery; (2) patients who have end-stage renal disease (such as glomerular filtration rate <15mL / (min·1.73m2), chronic dialysis or a history of kidney transplantation) at the time of admission; (3) clinical data are missing; (4) death within 24 hours after surgery. If it is determined based on the electronic medical record data that a patient meets at least one of the above criteria, the electronic medical record data of the patient is excluded.

[0028] Optionally, the predictor variable is used to predict renal dialysis, and the number of the predictor variables can be 39, which are preoperative variables or intraoperative variables. In addition, if the patient has multiple biochemical index measurement values ​​before surgery in the historical case information (if there are multiple data corresponding to the same predictor variable), the value of the patient when he was just hospitalized is used.

[0029] Optionally, indications for renal dialysis include symptoms such as uremia, oliguria or anuria for >2 days, fluid overload, or biochemical abnormalities (acidosis, hyperkalemia, BUN ≥29mmol / L, Scr ≥530μmol / L).

[0030] Optionally, clinical features are screened from the predictor variables according to preset screening conditions, including: eliminating predictor variables with variance inflation factors greater than a first preset value and correlation coefficients greater than a second preset value to obtain a second variable; obtaining clinical features from the second variable according to a preset method, the preset method including a LASSO regression method, preoperative features in the clinical features including at least one of serum creatinine, platelets, cardiovascular surgery history, right coronary artery involvement, and terminal site of dissection involvement, and intraoperative features including at least one of urine volume and estimated blood loss. The preset screening conditions are used to obtain clinical features with high consistency, non-redundancy, and the highest correlation with the prediction of renal dialysis.

[0031] Optionally, before screening clinical features by pre-set screening conditions, predictor variables with missing values ​​greater than a pre-set percentage can be excluded. For missing values ​​of other predictor variables, random forest method is used for filling.

[0032] In one embodiment, by eliminating the predictor variables with missing values ​​greater than 10%, the remaining missing values ​​are filled using the random forest method. Then, collinearity analysis is used to reduce the impact of homogeneous variance on the prediction model, and predictor variables with variance inflation factor>10 are eliminated. The eliminated predictor variables include age, left ventricular ejection fraction, hemoglobin, blood glucose, estimated glomerular filtration rate, extracorporeal circulation time, lowest body temperature, aortic clamping time, aortic arch replacement and operation time. In addition, in order to reduce the model error caused by the mutual influence between predictor variables, predictor variables with Pearson absolute correlation coefficient>0.5 are eliminated. Finally, the LASSO regression method is used to select the optimal predictor variables as the input variables of the model.

[0033] Optionally, a preset classifier is trained according to historical clinical feature information to obtain a prediction model, including: obtaining historical clinical feature information from the historical case information according to clinical characteristics, and forming a training cohort and a test cohort based on the historical clinical feature information; training and verifying a preset classifier according to the training cohort to obtain a prediction model.

[0034] Optionally, the historical clinical characteristic information is divided in a random manner to obtain a training cohort and a test cohort.

[0035] In one embodiment, the ratio of the amount of data of the training cohort to the test cohort is 4: 1. In addition, in the process of training the classifier, in order to determine the optimal classifier for predicting renal dialysis, a bootstrapping algorithm with different random seeds is further used, 80% of the training cohort is used for training, and 20% of the training cohort is used for verification, and the process is repeated 10 times. In different training processes, the historical clinical feature information included in the training cohort used for training is different.

[0036] Optionally, determining a target prediction model based on performance parameters of each prediction model includes: obtaining performance parameters of each test cohort based on a test cohort and a training cohort, the performance parameters including at least one of discrimination, calibration, and clinical utility evaluation.

[0037] Optionally, the discrimination is determined by the following parameters: AUC (Area Under Curve, the area under the ROC curve and the coordinate axis), accuracy, sensitivity, specificity, PPV (positive predictive value), NPV (negative predictive value) and F1 score (also called F1-Score or F1-Measure, which is a comprehensive indicator for evaluating the performance of a classification model. It is the harmonic mean of precision and recall) of the prediction model.

[0038] In one embodiment, when calculating the AUC value, the DeLong test can be used to compare whether the difference in AUC values ​​between different classifiers is statistically significant (if the difference is less than a preset difference value (which can be 0.05), it is determined to be statistically significant), and the optimal cutoff value is determined according to the maximum Youden index. The target prediction model is obtained using the classifier with the best prediction performance. The training and validation of the target prediction model are performed on the training cohort (80% of the data set) with 10-fold cross validation, and the cross validation process is repeated 10 times. The parameters of the target prediction model are manually adjusted on the basis of grid search, and the validation cohort is used for calibration, and the indicator of model optimization is accuracy.

[0039] Optionally, after the target prediction model is obtained through the training cohort, it can also be tested on the test cohort (accounting for 20% of the entire dataset). The performance of the target model is evaluated from the following three aspects: (1) The discrimination of the model is measured by AUC, accuracy, sensitivity, specificity, PPV, NPV and F1 score; (2) The calibration of the model is evaluated using the Brier score (defined as the average of the squared differences between the predicted outcome and the actual outcome). When the Brier score is ≤0.25, the model is considered to have good calibration; (3) The clinical utility of the model is evaluated by decision curve analysis.

[0040] In one embodiment, the prediction performance of different classifiers in the training cohort and the validation cohort is shown in Table 1. Among the six machine learning classifiers, RF showed the best discrimination ability, with an AUC of 0.864 (95% CI: 0.814-0.913) in the training cohort and an AUC of 0.715 (95% CI: 0.814-0.913) in the validation cohort. In addition, RF achieved the highest sensitivity of 0.764 (95% CI: 0.722-0.806) in the training cohort and remained stable at 0.735 (95% CI: 0.665-0.805) in the validation cohort. Therefore, we selected RF to obtain the target prediction model.

[0041] Table 1

[0042]

[0043]

[0044]

[0045] As shown in Table 1, in the training cohort, the prediction model obtained by RF had an AUC of 0.864 (95% CI: 0.818-0.910), a sensitivity of 0.767 (95% CI: 0.737-0.797), a specificity of 0.814 (95% CI: 0.777-0.852), and an accuracy of 0.803 (95% CI: 0.777-0.830). When the model performance was tested by the validation cohort and the test cohort, its discrimination remained good, with an AUC of 0.712 (0.519-0.904), a sensitivity of 0.788 (0.698-0.877), a specificity of 0.669 (0.562-0.776), and an accuracy of 0.718 (0.661-0.775) in the validation cohort. Moreover, the model had an AUC of 0.763, a sensitivity of 0.786, a specificity of 0.817, and an accuracy of 0.773 in the test cohort (not shown in the table). The Brier score was 0.163, demonstrating good calibration of the model. When evaluated by clinical utility, it provided better clinical net benefit compared with the "all treatment" or "no treatment" approach.

[0046] S101: Obtain clinical information of the predicted object.

[0047] Optionally, the clinical information can be obtained by manual input or by querying information such as the hospitalization number, medical card number, etc. of the predicted object.

[0048] The object for acquiring the clinical information may be an application encapsulating a target prediction model (such as a dialysis risk calculator), which receives the clinical information of the predicted object and uses the clinical information to calculate the risk of the predicted object undergoing renal dialysis.

[0049] Optionally, the prediction object is a subject undergoing surgery, and the surgery may be a cardiovascular surgery. The clinical information includes information corresponding to clinical characteristics, and the clinical characteristics include preoperative characteristics and intraoperative characteristics.

[0050] Optionally, the prediction object can be a patient undergoing surgical treatment for acute type A aortic dissection, or a patient who is prone to renal dialysis after other surgical treatments.

[0051] Optionally, clinical characteristics are characteristics that are closely associated with the risk of renal dialysis. Preoperative characteristics include 5 preoperative variables (serum creatinine, platelets, history of cardiovascular surgery, right coronary artery involvement, terminal site of dissection involvement) and intraoperative characteristics include 2 intraoperative variables (urine output and estimated blood loss), and these 7 variables are significantly associated with the risk of renal dialysis.

[0052] Among them, the relevance or importance of the above clinical characteristics to the risk of renal dialysis can be explained by the SHAP algorithm. The algorithm can calculate the impact of each clinical feature on the prediction result. Serum creatinine, platelets and terminal site of dissection are the three most important features, followed by urine volume, estimated blood loss, history of cardiovascular surgery and right coronary artery involvement.

[0053] S102: Inputting clinical information into a target prediction model, and obtaining the renal dialysis risk of the predicted object according to output information of the target prediction model.

[0054] Optionally, the target prediction model is a model trained according to historical medical record information in the above embodiment. Specifically, the target prediction model is trained by an RF classifier.

[0055] Optionally, obtaining the renal dialysis risk of the predicted object according to the output information of the target prediction model includes: displaying the probability of the predicted object to have renal dialysis according to the output information, and displaying the correlation between different clinical characteristics and the probability according to the output information. The correlation can be displayed through the prediction process, and the clinical characteristics that play a major role in the occurrence of renal dialysis are shown through the correlation.

[0056] In one embodiment, the target prediction model can be encapsulated in an application (such as a dialysis risk calculator for patients undergoing surgery for acute type A aortic dissection), which receives the numerical values ​​of the seven clinical characteristics of the prediction object input by the user, and calculates the probability of renal dialysis in a short time (within a few seconds). In addition, the prediction process is displayed to facilitate the user to understand which clinical characteristics play a major role in the prediction.

[0057] The present invention uses an interpretable target prediction model as a support tool for the management of postoperative renal dialysis in patients undergoing surgery for acute type A aortic dissection. The present invention screened out a total of 7 preoperative and intraoperative clinical factors that were significantly associated with the risk of renal dialysis. After comparing the performance of several commonly used classifiers in predicting renal dialysis, it was found that the RF classifier achieved the best discrimination, with good calibration and clinical practicality. Finally, a simple, interpretable, web-based risk calculator was developed based on the random forest classifier and the SHAP algorithm to facilitate clinical application.

[0058] If renal dialysis after major cardiovascular surgery is predicted based on preoperative variables alone, without considering the individual patient's physiological response to the surgical procedure and the occurrence of intraoperative adverse events, the risk of renal dialysis may be underestimated. Reassessment is a useful procedure when the patient is admitted to the intensive care unit after surgery, which complements preoperative risk stratification and helps formulate treatment strategies to reduce the risk of postoperative complications. Therefore, both preoperative and intraoperative variables should be used to build machine learning models. In addition, aortic dissection is a serious disease involving large vessels, and vascular status has a very important impact on the development of patient outcomes. Therefore, the patient's preoperative vascular characteristics should also be included to show the involvement of the dissection. Finally, five preoperative variables (serum creatinine, platelets, history of cardiovascular surgery, right coronary artery involvement, and terminal site of dissection involvement) and two intraoperative variables (urine output and estimated blood loss) were significantly associated with the risk of renal dialysis.

[0059] Patients with acute type A aortic dissection are more likely to require renal dialysis. As can be seen from our study, this population often has longer intensive care hospital stays, longer mechanical ventilation times, higher rates of tracheal reintubation, and even higher mortality rates. Therefore, this dilemma has placed a huge financial burden on an already challenged healthcare system. Risk assessment of renal dialysis can help with early intervention and improve patient outcomes. If the method of the present application determines that the risk of a subject undergoing renal dialysis is higher than the preset probability (e.g., greater than 70%), the following methods can be taken subsequently: (1) avoid the use of nephrotoxic drugs before, during, and after surgery; (2) optimize patient hemodynamics and use positive inotropic drugs to improve renal perfusion; (3) closely monitor body fluid and renal status; and (4) control perioperative blood glucose levels.

[0060] The method of this application can correctly identify high-risk patients, thereby guiding the intensive care unit to allocate postoperative care resources in an effective manner. The prediction method of this application will help clinicians easily understand the logic behind the decision, allow patients and their families to obtain comprehensive informed consent, and help identify high-risk subjects for clinical trials of renal protection treatments.

[0061] In an optional embodiment, a renal dialysis prediction device is provided, such as Figure 2 As shown, the prediction device includes a clinical information acquisition module and a prediction module. The clinical information acquisition module is used to obtain clinical information of the prediction object, and the clinical information includes information corresponding to clinical characteristics, and the clinical characteristics include preoperative characteristics and intraoperative characteristics; the prediction module is used to input the clinical characteristic information into the target prediction model, and obtain the renal dialysis risk of the prediction object according to the output information of the target prediction model, and the target prediction model is obtained by training a preset classifier based on historical medical record information.

[0062] Optionally, multiple preset classifiers are trained based on historical medical record information, including: obtaining historical clinical feature information based on the historical medical record information of historical patients; training multiple preset classifiers based on the historical clinical feature information to obtain a prediction model, and determining a target prediction model based on the performance parameters of each prediction model, the preset classifiers including at least two of LightGBM, XGBoost, SVM, RF, AdaBoost and Logistic regression.

[0063] Optionally, clinical characteristic information is obtained based on historical medical records of historical patients, including: processing the historical medical records information based on a first preset standard to generate training data, the first preset standard including at least one of treatment time, disease type, degree of completeness of clinical data, and postoperative status; obtaining predictive variables corresponding to the training data, and screening clinical characteristics from the predictive variables according to preset screening conditions, the preset screening conditions including at least one of high consistency, non-redundancy, and correlation.

[0064] Optionally, clinical characteristics are screened from the predictor variables according to preset screening conditions, including: eliminating predictor variables with a variance inflation factor greater than a first preset value and a correlation coefficient greater than a second preset value to obtain a second variable; obtaining clinical characteristics from the second variable according to a preset method, the preset method including a LASSO regression method, the preoperative characteristics in the clinical characteristics including at least one of serum creatinine, platelets, history of cardiovascular surgery, right coronary artery involvement, and terminal site of dissection involvement, and the intraoperative characteristics including at least one of urine volume and estimated blood loss.

[0065] Optionally, a preset classifier is trained according to the historical clinical feature information to obtain a prediction model, including: obtaining historical clinical feature information from the historical case information according to the clinical characteristics, and forming a training cohort and a test cohort based on the historical clinical feature information; training and verifying the preset classifier according to the training cohort to obtain a prediction model.

[0066] Optionally, determining a target prediction model based on the performance parameters of each prediction model includes: obtaining the performance parameters of each test cohort based on the test cohort and the training cohort, the performance parameters including at least one of discrimination, calibration, and clinical utility evaluation.

[0067] Optionally, obtaining the renal dialysis risk of the predicted object according to the output information of the target prediction model includes: displaying the probability of the predicted object developing renal dialysis according to the output information, and displaying the correlation between different clinical characteristics and the probability according to the output information.

[0068] In an optional embodiment, an electronic device is provided, such as Figure 3 As shown, Figure 3 The electronic device 4000 shown includes: a processor 4001 and a memory 4003. The processor 4001 and the memory 4003 are connected, such as through a bus 4002. Optionally, the electronic device 4000 may also include a transceiver 4004, which may be used for data interaction between the electronic device and other electronic devices, such as data transmission and / or data reception. It should be noted that in actual applications, the transceiver 4004 is not limited to one, and the structure of the electronic device 4000 does not constitute a limitation on the embodiments of the present application.

[0069] Processor 4001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. It may implement or execute various exemplary logic blocks, modules and circuits described in conjunction with the disclosure of this application. Processor 4001 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0070] The bus 4002 may include a path for transmitting information between the above components. The bus 4002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus 4002 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0071] The memory 4003 may be a ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compressed optical disk, laser disk, optical disk, digital versatile disk, Blu-ray disk, etc.), magnetic disk storage media, other magnetic storage devices, or any other medium that can be used to carry or store computer programs and can be read by a computer, without limitation herein.

[0072] The memory 4003 is used to store the computer program for executing the embodiment of the present application, and the execution is controlled by the processor 4001. The processor 4001 is used to execute the computer program stored in the memory 4003 to implement the steps shown in the above method embodiment.

[0073] The electronic device may be any electronic product that can interact with an object, such as a personal computer, a tablet computer, a smart phone, a personal digital assistant (PDA), a game console, an interactive network television (IPTV), a smart wearable device, etc.

[0074] The electronic device may also include a network device and / or an object device, wherein the network device includes, but is not limited to, a single network server, a server group consisting of multiple network servers, or a cloud consisting of a large number of hosts or network servers based on cloud computing.

[0075] The network where the electronic device is located includes but is not limited to the Internet, a wide area network, a metropolitan area network, a local area network, a virtual private network (VPN), etc.

[0076] An embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps and corresponding contents of the aforementioned method embodiment can be implemented.

[0077] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0078] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for predicting renal dialysis, characterized in that: The method comprises: S101: Acquire clinical information of the predicted object, wherein the clinical information includes information corresponding to clinical characteristics, wherein the clinical characteristics include preoperative characteristics and intraoperative characteristics, wherein the predicted object is a patient with acute type A aortic dissection, wherein the preoperative characteristics include serum creatinine, platelets, cardiovascular surgery history, right coronary artery involvement, and dissection final uninvolved site, and the intraoperative characteristics include urine volume and estimated blood loss; S102: inputting the clinical information into a target prediction model, and obtaining the renal dialysis risk of the predicted object according to the output information of the target prediction model, wherein the target prediction model is trained by an RF classifier based on historical medical record information; The target prediction model is trained by the RF classifier based on historical medical record information, including: Obtain historical clinical characteristic information based on historical medical record information of historical patients; The target prediction model is obtained by training the RF classifier according to the historical clinical feature information.

2. The method for predicting renal dialysis according to claim 1, wherein: The step of obtaining the renal dialysis risk of the predicted object according to the output information of the target prediction model comprises: The probability of the predicted object requiring renal dialysis is displayed according to the output information, and the correlation between different clinical characteristics and the probability is displayed according to the output information.

3. A renal dialysis prediction device, characterized in that: include: A clinical information acquisition module is used to acquire clinical information of the predicted object, wherein the clinical information includes information corresponding to clinical characteristics, wherein the clinical characteristics include preoperative characteristics and intraoperative characteristics, wherein the predicted object is a patient with acute type A aortic dissection, wherein the preoperative characteristics include serum creatinine, platelets, cardiovascular surgery history, right coronary artery involvement, and the final uninvolved part of the dissection, and the intraoperative characteristics include urine volume and estimated blood loss; A prediction module, used for inputting the clinical feature information into a target prediction model, and obtaining the renal dialysis risk of the predicted object according to the output information of the target prediction model, wherein the target prediction model is trained by an RF classifier based on historical medical record information; The target prediction model is trained by the RF classifier based on historical medical record information, including: Obtain historical clinical characteristic information based on historical medical record information of historical patients; The target prediction model is obtained by training the RF classifier according to the historical clinical feature information.

4. An electronic device, characterized in that: The electronic device includes a processor and a memory, wherein the memory stores a computer program, and the computer program is used by the processor to implement the renal dialysis prediction method according to claim 1 or 2.

5. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program data, and the program data is used to execute the renal dialysis prediction method according to claim 1 or 2.