A diagnostic prediction model for liver cirrhosis combined with portal vein thrombosis and a method for establishing the same

Abstract

The present application relates to the field of biotechnology, in particular to a diagnosis prediction model for liver cirrhosis combined with portal vein thrombosis and a method for establishing the same, wherein the method for establishing the same screens 7 risk factors from blood data and endoscopic operation parameters of patients, including splenectomy history, blood urea nitrogen, Na, activated partial thromboplastin time, thrombin time, D-dimer and the degree of esophageal varices under endoscopy. A nomogram model is established based on the 7 risk factors, and the area under the ROC curve (AUC) is 0.907 (95% CI, 0.877-0.931). The calibration curve, clinical decision curve and clinical impact curve show that the model has good prediction accuracy. The AUC of the external validation queue of 63 patients is 0.948 (95% CI, 0.860-0.988). It can be used to accurately predict the occurrence of portal vein thrombosis, and has a broad application prospect in clinical practice.

Description

Technical Field

[0001] This invention relates to the field of biotechnology, specifically to a diagnostic prediction model for cirrhosis complicated with portal vein thrombosis and its establishment method. Background Technology

[0002] In cirrhosis, especially in the decompensated stage, the body's coagulation function undergoes complex changes, altering procoagulant and anticoagulant substances, ultimately leading to bleeding or thrombosis. Portal vein thrombosis (PVT) refers to the formation of a thrombus within the main portal vein, which can extend to intrahepatic or extrahepatic venous branches. PVT is a major and serious complication of cirrhosis. The structural disturbances in the liver caused by cirrhosis and portal hypertension, slowed portal vein blood flow, and damage to the vessel wall induce hypercoagulability, playing a crucial role in the development of PVT. The prevalence of PVT in cirrhosis ranges from 0.6% to 16%. Cirrhosis complicated by PVT can worsen liver function; PVT patients sometimes experience only mild abdominal pain or are even asymptomatic. However, if the thrombus obstructs the main portal vein and its branches, it can lead to fatal complications such as massive gastrointestinal bleeding, cavernous degeneration of the portal vein, intestinal ischemia and necrosis, multiple organ failure, and even death. Therefore, timely and accurate diagnosis is essential.

[0003] Since pathological features from physical examination and laboratory tests cannot be used to diagnose portal vein thrombosis (PVT), technicians primarily rely on imaging techniques. Doppler ultrasound is the preferred imaging technique for screening PVT, as it can simultaneously observe the location of the thrombus and the direction and velocity of blood flow in the portal vein. A portal vein flow velocity <15 cm / s can predict the occurrence of PVT, but this can be affected by intestinal distension or obesity. Incomplete occlusion of the PVT, thrombosis only in the superior mesenteric vein, or portal vein thrombosis confined to the posterior duodenum can all render Doppler ultrasound results unreliable. Secondly, PVT in cirrhosis often needs to be differentiated from portal vein tumor thrombosis caused by primary or metastatic liver cancer. CT and MRI are currently the best methods for distinguishing between thrombi and tumor thrombi, but they also have drawbacks such as high cost and inconsistent quality.

[0004] Researchers in this field have developed a predictive model for portal vein thrombosis in patients with cirrhosis (CN116313106A) using baseline patient data. The model's independent predictive features include NLR, PLR, GLR, portal vein width, APRI, spleen subcostal distance, and D-dimer; the area under the AUC is 0.841. However, this predictive model still suffers from low specificity and accuracy, leading to misdiagnosis.

[0005] To address the aforementioned technical problems, this invention provides a method for establishing a diagnostic prediction model for cirrhosis complicated with portal vein thrombosis. The method used to screen predictive features such as BUN, Na, APTT, TT, D-dimer, degree of esophageal and gastric varices, and splenectomy were obtained. In the obtained prediction model, the AUC value was 0.907 (95% CI, 0.877–0.931), and the p-value of the De Long test was 0.001, achieving unexpected beneficial results. Summary of the Invention

[0006] The primary objective of this invention is to provide a method for establishing a diagnostic prediction model for cirrhosis complicated with portal vein thrombosis. The method for establishing the prediction model is as follows:

[0007] (1) Collect variables from cirrhotic patients undergoing endoscopy and identify patient baseline data, blood data and endoscopic operation parameters as predictive factors;

[0008] (2) Based on the baseline data obtained in step (1), patients were matched with a 1:1 propensity score and divided into an experimental group and a control group.

[0009] (3) The blood data and endoscopic operation parameters of patients in the experimental group and the control group were excluded from collinearity by machine learning algorithm, and risk factors were screened by RF algorithm;

[0010] (4) The risk factors mentioned in step (3) were identified and analyzed by univariate and multivariate logistic regression analysis, and risk factors with p < 0.05 were selected for modeling;

[0011] (5) Establish a nodal diagram model using the risk factors described in step (4);

[0012] (6) After the nomogram model described in step (5) is established, the model is validated using an external validation group.

[0013] Preferably, the patient baseline information described in step (1) includes: age, sex, BMI, Child-Pugh classification of liver function, etiology and type of cirrhosis, hypertension, history of diabetes, smoking, drinking history, family history, splenectomy, and history of previous endoscopic treatment.

[0014] Preferably, the blood data in step (1) includes: coagulation indicators, biochemical indicators, and routine blood test data including red blood cell count, white blood cell count, platelet count, platelet distribution width, and hemoglobin; the endoscopic operation parameters include: degree of esophageal and gastric varices, red signs, whether there is bleeding, and history of endoscopic treatment.

[0015] Preferably, the machine learning method for excluding multicollinearity of variables in step (3) is ridge regression.

[0016] Preferably, the RF algorithm used in step (3) to screen for risk factors is as follows: the RF parameters are optimized around their default values ​​in a logarithmic step, wherein the RF parameters are 500 trees and a random subspace, and the dimension is equal to the rounded value of the square of the number of features.

[0017] Preferably, in step (4), tenfold cross-validation and an internally set test set are used to verify the reliability of the hazard factors.

[0018] Preferably, the risk factors obtained in step (5) are history of splenectomy, blood urea nitrogen (BUN), sodium, activated partial thromboplastin time (APTT), thrombin time (TT), D-dimer, and the degree of endoscopic esophageal varices.

[0019] A second objective of this invention is to provide a predictive model for cirrhosis complicated with portal vein thrombosis established by the method described.

[0020] A third object of the present invention is to provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the system performs the following steps:

[0021] (1) Collect and / or input variable data of patients with cirrhosis complicated with portal vein thrombosis;

[0022] (2) Substitute the variable data into the model for calculation to obtain the judgment conclusion on whether cirrhosis is complicated with portal vein thrombosis;

[0023] (3) Output the conclusion on whether cirrhosis is complicated by portal vein thrombosis.

[0024] A fourth object of the present invention is to provide a computer-readable storage medium having a computer device as described thereon stored thereon.

[0025] The beneficial effects of this invention are as follows: This invention provides a method for establishing a diagnostic prediction model for portal vein thrombosis in patients with cirrhosis. The method identifies seven risk factors from patient blood data and endoscopic operation parameters, including a history of splenectomy, blood urea nitrogen (BUN), sodium (Na), activated partial thromboplastin time (APTT), thrombin time (TT), D-dimer, and the degree of endoscopic esophageal varices. A nomogram model was established using these seven risk factors, with an area under the ROC curve (AUC) of 0.907 (95% CI, 0.877–0.931). Calibration curves, clinical decision curves, and clinical impact curves show that the model has good predictive accuracy. The AUC of a 63-patient external validation cohort was 0.948 (95% CI, 0.860–0.988). This model can be used to accurately predict the occurrence of portal vein thrombosis and has broad clinical application prospects. Attached Figure Description

[0026] Figure 1 This shows a flowchart of the patient recruitment process for this study.

[0027] Figure 2 Random forest for selecting variables

[0028] Figure 3 Evaluate the predictive power of the integrated model

[0029] Note: Ⅰ: ROC curve; Ⅱ: Calibration curve; Ⅲ: DCA curve; Ⅳ: CIC curve

[0030] Figure 4 Evaluate the predictive power of the integrated model in the external validation queue.

[0031] Note: Ⅰ: ROC curve; Ⅱ: Calibration curve; Ⅲ: DCA curve; Ⅳ: CIC curve

[0032] Figure 5 Schematic diagram of line chart model Detailed Implementation

[0033] This invention provides a method for constructing a prediction model for portal vein thrombosis in patients with cirrhosis and a nomogram model. To make the objectives, technical solutions, and effects of this invention clearer and more explicit, the invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0034] In the description of this invention, it should be understood that the terms "installation" and "connection" should be interpreted broadly, and those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0035] SPSS v.22.0 (IBM, Armonk, New York, USA) was used to analyze the data. Mean ± standard deviation was used to describe normally distributed data and variables. The median (interquartile range) was used to describe data that did not conform to a normal distribution. Frequency (percentage) was used to describe the categorical distribution of variables. Differences between groups were compared using t-tests, chi-square tests, or rank-sum tests.

[0036] The Child-Pugh liver function classification is generally divided into three grades: A, B, and C, with different specific criteria. A grade A score of 5-6 indicates strong liver compensatory capacity, lower surgical risk, and a good prognosis, with a 1-2 year survival rate of 85%-100%. A grade B score of 7-9 indicates some liver compensatory capacity, moderate surgical risk, and a relatively poor prognosis, with a 1-2 year survival rate of 60%-80%. A grade C score of 10-15 typically indicates decompensated liver function, higher surgical risk, and a 1-2 year survival rate of 35%-45%.

[0037] Classification Scoring Criteria Surgical risk 1-2 year survival rate Grade A 5-6 points smaller 85％-100％ Grade B 7-9 points moderate 60％-80％ Class C 10-15 points Larger 35％-45％

[0038] Example 1: A diagnostic prediction model for cirrhosis complicated by portal vein thrombosis

[0039] 1. Research Subjects

[0040] This retrospective study was conducted at the Endoscopy Center of Gansu Provincial People's Hospital, in accordance with the Declaration of Helsinki, and approved by the Ethics Committee of the Medical Department of Gansu Provincial People's Hospital (Ethics No.: 2023-361). Written informed consent was obtained from patients prior to the endoscopy. The model was based on the TRIPOD guidelines.

[0041] Patients with cirrhosis who underwent endoscopy between October 2019 and January 2023 were included. Exclusion criteria: patients with liver tumors, liver transplant recipients, Budd-Chiari syndrome, patients receiving TIPS or surgical treatment, patients receiving anticoagulation therapy, patients with severe disease, and patients with missing data.

[0042] A total of 1058 eligible patients were initially screened. Exclusion criteria included 148 patients with concurrent tumors, 78 patients treated with TIPS or surgery, 74 patients on anticoagulation therapy, 67 patients with severe illness, and 42 patients with insufficient data. A final enrollment of 649 eligible patients was made. Propensity score matching (PSM) was performed 1:1 based on baseline characteristics prior to endoscopy.

[0043] Sixty-three eligible patients were enrolled between January and May 2023 for external validation.

[0044] 2. Data Collection

[0045] The data were collected independently by two researchers. The top 54 indicators for ultrasound diagnosis were screened for analysis: (1) Patient baseline information included: age, gender, BMI, Child-Pugh liver function, etiology and type of cirrhosis, hypertension, history of diabetes, smoking, drinking history, family history, splenectomy, and history of previous endoscopic treatment; (2) Blood data included: coagulation indicators, biochemical indicators, and routine blood test data including red blood cell count (RBC), white blood cell count (WBC), platelet count (PLT), platelet distribution width (PDW), and hemoglobin (HGB); (3) Endoscopic parameters included: degree of esophageal and gastric varices, red sign, bleeding, and history of endoscopic treatment.

[0046] 3. Diagnostic criteria for PVT

[0047] In cases of cirrhosis complicated by portal vein thrombosis, color Doppler ultrasound or multi-slice spiral computed tomography (CT), hepatic angiography, and liver-specific magnetic resonance imaging show that the portal vein or its branches are basically unclear or have interrupted blood flow, with filling defects or even disappearance.

[0048] 4. Establishment of the prediction model

[0049] (1) Based on the patient baseline data in the materials collected in 2, patients were matched 1:1 with propensity score and divided into experimental group and control group. Chi-square test and t test were used.

[0050] (2) Risk factors were screened from the blood test data and endoscopic features obtained in step 2 using machine learning methods; ridge regression was used to exclude multicollinearity of variables. Random forest (RF) was used to identify the main candidate features and select the appropriate program based on its impact on the prediction results. The RF parameters were optimized around their default values ​​in logarithmic steps (using 500 trees and a random subspace whose dimension is equal to the rounded value of the square of the number of features). The machine learning methods were implemented using R (version 4.1.1).

[0051] (3) Use tenfold cross-validation and internal test set validation to verify the reliability of the modeling variables selected in step (2);

[0052] (4) Candidate factors for the training group RF algorithm are determined by univariate and multivariate logistic regression analysis. The risk factors obtained in step (3) are incorporated into the binary logistic regression model in univariate and multivariate form. The risk factors are analyzed and variables with p < 0.05 are selected.

[0053] (5) Use the variables obtained in step (4) to build a noctilinear plot model;

[0054] (6) After the nomogram model described in step (5) is established, the model is validated using an internal validation group and an external validation group.

[0055] The accuracy of the predictive model was evaluated using an external validation set from Gansu Provincial People's Hospital from January to May 2023. ROC, DCA, and CIC curve analyses were employed, and the model's accuracy was verified by estimating the difference between the validation set and the nomogram.

[0056] A p-value less than 0.05 was considered statistically significant. The adjusted odds ratio (OR) and the corresponding 95% confidence interval (95% CI) were calculated. The receiver operating characteristic (ROC) was estimated, including the area under the curve (AUC) and the Hosmer-Lemeshow test. The accuracy was tested by DCA (decision curve analysis) curve and the predicted probability was compared with the actual probability by CIC (clinical influence curve) curve.

[0057] 5. Results

[0058] (1) Patient characteristics

[0059] A total of 649 eligible patients were included. Figure 1 Regarding baseline characteristics before endoscopy, there were no differences between the PVT group and the control group in terms of gender, type of cirrhosis, history of hypertension, history of diabetes, smoking history, alcohol consumption history, family history, and splenectomy. However, there were statistically significant differences between the two groups in baseline aspects such as age, BMI, history of endoscopic treatment, and Child-Pugh classification before endoscopy (Table 1). No differences were found after PSM 1:1 matching. A total of 470 patients were ultimately included through PSM 1:1 matching. In addition, 63 out-of-group patients were included from January 2023 to May 2023 to validate the accuracy of the model.

[0060] Table 1 Analysis of patient baseline data

[0061]

[0062]

[0063] (2) Variable selection

[0064] We used ridge regression to select feature variables from the included data. All 76 variables were excluded from collinearity and included in the selection variables using the random forest (RF) method. To obtain the optimal feature set, the importance of each variable was calculated; finally, 30 feature variables were identified using the random forest (RF) method, such as... Figure 2 As shown.

[0065] (3) Logistic analysis prediction model

[0066] Candidate factors for the RF algorithm in the training group were identified through univariate and multivariate logistic regression analysis. In the PVT group, BUN (bunion urinary tract infection) was a risk factor for portal vein thrombosis (BUN > 140.1 mmol / L) (OR = 1.15795% CI 1.247–1.306).

[0067] Serum sodium levels not only reflect the degree of liver damage but are also a serological indicator closely related to liver damage. Results showed that hyponatremia was significantly associated with prognosis, with severe hyponatremia indicating a poor prognosis. Hepatorenal syndrome is a serious decompensated complication of cirrhosis, characterized by severe hyponatremia and azotemia. This study found that serum sodium level was a protective factor against portal vein thrombosis (Na < 140.1 mmol / L) (OR = 0.910, 95% CI 0.848–0.977).

[0068] The liver plays a crucial role in the hemostatic system because it synthesizes most of the coagulation factors and proteins involved in fibrinolysis, and it also synthesizes thrombopoietin, which is responsible for platelet production. Due to reduced vitamin K synthesis in patients with cirrhosis, the synthesis of vitamin K-related coagulation factors II, V, VII, IX, and X is decreased, thus affecting the intrinsic coagulation pathway. The length of the acute prothrombin time (APTT) primarily reflects the quality and quantity of intrinsic coagulation factors and can serve as an indicator of liver synthetic function. Prothrombin time reflects platelet count and vascular wall function. With the progression of cirrhosis, the endothelial cell procoagulant phenotype has shown that upregulation of certain markers is an important risk factor for portal vein thrombosis (PVT). Many studies have shown that P-selectin and vWF are upregulated in dysfunctional portal vein endothelial cells, a contributing factor to PVT. Severe portal hypertension increases the damage to the vascular endothelial cell wall and also increases intimal hyperplasia. In our study, we found that the optimal cutoff value for APTT time was >32.7 s (OR = 1.089; 95% CI, 1.039–1.140), and the optimal critical value was TT time <16.4 s (OR = 0.703; 95% CI, 0.606–0.814), with a sensitivity of 51.5% and a specificity of 66.7%.

[0069] D-dimer is a small peptide fragment present in the early stages of thrombosis, playing a role in fibrinogen degradation and plasmin formation. Its diagnostic value in thrombotic events has been established. D-dimer can predict the occurrence of thrombotic events (PVT) and serves as a stable and reliable indicator. In this study, D-dimer >0.87 ug / ml was an independent risk factor (DD > 0.87 g / L) (OR = 1.925, 95% CI, 1.538–2.410).

[0070] Moderate to severe esophageal and gastric varices (OR = 2.792; 95% CI, 1.577–4.944), with a sensitivity of 94.7%, can directly reflect the degree of portal hypertension in cirrhosis. The degree of gastric varices is correlated with the severity of cirrhosis. Patients with PVT have elevated portal vein pressure, which exacerbates the severity of esophageal and gastric varices. One study found that moderate to severe esophageal and gastric varices observed endoscopically are an independent risk factor. Previous studies have shown that the presence of esophageal and gastric varices is significantly associated with the incidence of PVT, indicating that esophageal and gastric varices may be an important part of assessing PVT risk. In this study, esophageal and gastric varices were considered an important indicator for predicting PVT in patients with cirrhosis. Gastroscopy is a routine examination for patients with cirrhosis; therefore, when moderate to severe esophageal and gastric varices are found during gastroscopy, the possibility of PVT should be ruled out in conjunction with other predictive factors.

[0071] Following splenectomy, hemodynamic changes occur in the portal venous system, with reduced portal venous blood flow and velocity, and decreased platelet destruction, leading to a significant increase in platelet count in peripheral vessels. In most cases, platelet, red blood cell, and white blood cell counts significantly increase shortly after splenectomy in patients with cirrhosis, indicating a hypercoagulable state. Therefore, previous studies have shown that preoperative low platelet and white blood cell counts are precursors to postoperative portal vein thrombosis (PVT). One study found that the platelet elevation rate on postoperative day 1 (PPER1) and day 3 (PPER3) are effective indicators for predicting PVT after splenectomy. The PER-based model for predicting PVT has an accuracy of up to 80%. Due to mechanical damage to the portal vein, endothelial collagen fibers are exposed, and the coagulation system is activated. Therefore, patients after splenectomy are more prone to PVT. This study found that splenectomy is an independent risk factor for portal vein thrombosis (OR = 7.966, 95% CI 2.911–21.795).

[0072] High-risk factors for the comprehensive model of portal vein thrombosis are shown in Tables 2 and 3. A nomogram predicting portal vein thrombosis was prepared using the above seven risk factors.

[0073] The nomogram model combines endoscopic features of PVT with clinical blood test data. Through calibration and accuracy testing, the validated nomogram demonstrates high predictive value. Figure 3 As shown, the predictive model has an AUC of 0.907 (95% CI, 0.877–0.931) and a p-value of 0.001 for the De Long test, outperforming any single predictive model. The calibration curve represents the degree of agreement between the predicted and actual probabilities, and DCA represents the model's net clinical benefit. The accuracy was verified through DCA and CIC curve analysis.

[0074] Table 2 Univariate and multivariate logistic regression analyses of cirrhosis complicated with portal vein thrombosis.

[0075]

[0076] Table 3. Predicted values ​​of risk factors.

[0077] factor Cutoff value Sensitivity Specificity AUC BUN (mmol / L) >6.91 26.9 84 0.555 Na (mmol / L) <140.1 77.5 52.7 0.680 ATPP(S) >32.7 71.4 45.7 0.604 TT(S) <16.4 51.5 66.7 0.610 DD (g / L) >0.87 83.3 83.1 0.858 Degree of esophageal and gastric varices 1 94.7 38.7 0.708 History of splenectomy 1 45.4 90.1 0.677

[0078] (4) External validation of the model

[0079] Sixty-three patients were included (31 with PVT, 49.21%; 32 in the control group) to validate the nomogram of the training set, followed by ROC, DCA, and CIC curve analyses. The AUC value of the predicted ROC curve was 0.948 (95% CI, 860–0.988), with a sensitivity of 90.3% and a specificity of 90.6%. The calibrated curve, DCA curve, and CIC curve were all within the fitted range. Figure 4 ).

[0080] (5) Nodal plot of the model

[0081] For clinical applications, we have developed a public nomogram prediction tool. Users can predict PVT (Proportional Ventilation Test) using seven features. Figure 5 Each factor has a predictive reference value based on OR, which shows the weight of each parameter. The total score can distinguish between cirrhosis and cirrhosis combined with PVT.

[0082] In summary, this invention provides a method for establishing a diagnostic prediction model for cirrhosis complicated with portal vein thrombosis. The method identifies seven risk factors from patient blood data and endoscopic procedure parameters: history of splenectomy, blood urea nitrogen (BUN), sodium (Na), activated partial thromboplastin time (APTT), thrombin time (TT), D-dimer, and the degree of endoscopic esophageal varices. A nomogram model was established using these seven risk factors, with an area under the ROC curve (AUC) of 0.907 (95% CI, 0.877–0.931), sensitivity of 88.2%, and specificity of 85.3%. The p-value of the Hosmer-Lemeshow test was 0.919. The calibration curve, clinical decision curve, and clinical impact curve show that the model has good predictive accuracy. The AUC of the external validation cohort of 63 patients was 0.948 (95% CI, 0.860–0.988). It can be used to accurately predict the occurrence of portal vein thrombosis and has broad application prospects in clinical practice.

Claims

1. A method for establishing a predictive model for cirrhosis complicated with portal vein thrombosis, characterized in that, The method for establishing the prediction model is as follows: (1) Collect variables from cirrhotic patients undergoing endoscopy and identify patient baseline data, blood data and endoscopic operation parameters as predictive factors; (2) Based on the baseline data obtained in step (1), patients were matched with a 1:1 propensity score and divided into an experimental group and a control group. (3) The blood data and endoscopic operation parameters of patients in the experimental group and the control group were excluded from collinearity by machine learning algorithm, and risk factors were screened by RF algorithm; (4) The risk factors mentioned in step (3) were identified and analyzed by univariate and multivariate logistic regression analysis, and risk factors with p < 0.05 were selected for modeling; (5) Establish a nodal diagram model using the risk factors described in step (4); (6) After the nomogram model described in step (5) is established, the model is validated using an external validation group.

2. The method for establishing a predictive model for cirrhosis complicated with portal vein thrombosis as described in claim 1, characterized in that, The patient baseline information described in step (1) includes: age, sex, BMI, Child-Pugh liver function classification, type of cirrhosis etiology, hypertension, history of diabetes, smoking, drinking history, family history, splenectomy, and history of previous endoscopic treatment.

3. The method for establishing a predictive model for cirrhosis complicated with portal vein thrombosis as described in claim 1, characterized in that, The blood data mentioned in step (1) includes: coagulation indicators, biochemical indicators, and routine blood test data including red blood cell count, white blood cell count, platelet count, platelet distribution width, and hemoglobin; the endoscopic operation parameters include: degree of esophageal and gastric varices, red signs, whether there is bleeding, and history of endoscopic treatment.

4. The method for establishing a predictive model for cirrhosis complicated with portal vein thrombosis as described in claim 1, characterized in that, The machine learning method for excluding multicollinearity of variables described in step (3) is ridge regression.

5. The method for establishing a predictive model for cirrhosis complicated with portal vein thrombosis as described in claim 1, characterized in that, The RF algorithm used in step (3) to screen for risk factors is as follows: the RF parameters are optimized in logarithmic steps around their default values. The RF parameters are 500 trees and a random subspace, and their dimension is equal to the rounded value of the square of the number of features.

6. The method for establishing a predictive model for cirrhosis complicated with portal vein thrombosis as described in claim 1, characterized in that, In step (4), tenfold cross-validation and internally set test set validation are used to verify the reliability of the hazard factors.

7. The method for establishing a predictive model for cirrhosis complicated with portal vein thrombosis as described in claim 1, characterized in that, The risk factors obtained in step (4) are history of splenectomy, blood urea nitrogen (BUN), sodium, activated partial thromboplastin time (APTT), thrombin time (TT), D-dimer and the degree of endoscopic esophageal varices.

8. A predictive model for cirrhosis complicated with portal vein thrombosis established by the method described in any one of claims 1-7.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it causes the system of claim 7 to perform the following steps: (1) Collect and / or input variable data of patients with cirrhosis complicated with portal vein thrombosis; (2) Substitute the variable data into the model for calculation to obtain the judgment conclusion on whether cirrhosis is complicated with portal vein thrombosis; (3) Output the conclusion on whether cirrhosis is complicated by portal vein thrombosis.

10. A computer-readable storage medium having stored thereon the computer device as claimed in claim 9.

Images