A method for constructing a risk assessment model for multi-stage liver fibrosis

Abstract

The application discloses a kind of for the construction method of risk assessment model of multi-stage liver fibrosis, comprising the following steps: constructing observation sample set, observation sample set is made of independent variable and dependent variable, multi-stage liver fibrosis includes F1-F3 period and F4 period;Definition risk assessment model;Risk assessment model is used to evaluate the risk of multi-stage liver fibrosis according to serum TET3 concentration and FIB-4 index;Iterative calculation is carried out to risk assessment model, and the optimal stage coefficient corresponding to different evaluation stages is determined;The optimal stage coefficient is substituted into probability assessment model to generate probability assessment formula, and the probability value of each stage liver fibrosis is calculated according to the serum TET3 concentration and FIB-4 index to be evaluated, and the risk assessment of each stage liver fibrosis is realized in combination with probability threshold value.According to the above technical scheme, FIB-4 index and serum TET3 concentration can be combined to realize the risk assessment of liver fibrosis and cirrhosis, and the accuracy of evaluation can be improved in continuous application.

Description

Technical Field

[0001] This invention relates to the field of biomedical information processing technology, and more specifically, to a method for constructing a risk assessment model for multi-stage liver fibrosis. Background Technology

[0002] Controlling chronic liver disease is a major global public health issue, and liver fibrosis is a key pathological stage in the progression of chronic liver disease. Without timely intervention, liver fibrosis can further develop into cirrhosis and even hepatocellular carcinoma. Currently, liver biopsy is the "gold standard" for assessing the degree of liver fibrosis; however, this method is invasive, carries the risk of complications, and is not easily repeatable, limiting its application in large-scale screening and dynamic monitoring. Therefore, establishing an accurate, non-invasive, and convenient method for assessing liver fibrosis is of significant clinical importance.

[0003] Currently, commonly used non-invasive assessment methods for liver fibrosis in clinical practice fall into two main categories: serological markers and imaging techniques. Regarding serological markers, the FIB-4 (Fibrosis-4 index) is one of the most widely used indirect serological biomarkers. It can be calculated from indicators such as age and platelet count, offering the advantages of simplicity and ease of use; routine blood tests can complete the assessment. Multiple studies have confirmed that the FIB-4 index can be used for risk assessment of liver fibrosis caused by various etiologies, including chronic hepatitis B and non-alcoholic fatty liver disease. Currently, several auxiliary diagnostic techniques for liver fibrosis based on serological markers exist. For example, Chinese patent application CN117174296A discloses a diagnostic model for liver fibrosis that only requires testing the subject's ex vivo serum to obtain diagnostic results; Chinese patent application CN118430796A discloses a non-invasive diagnostic method and system for liver fibrosis based on blood biochemical indicators. However, the FIB-4 index lacks specificity in diagnosing advanced fibrosis and is insufficient for precise diagnosis when used alone. Furthermore, the FIB-4 index is affected by factors such as fluctuations in liver function and platelet counts, limiting its diagnostic accuracy. Meanwhile, research on novel biomarkers has revealed a significant effect of serum TET3 levels in patients with liver fibrosis; however, current technology lacks a method to combine serum TET3 and FIB-4 to construct a model for assessing the risk of liver fibrosis staging and to use it as an adjunct to the diagnosis of liver fibrosis.

[0004] Therefore, a technical solution is needed to achieve risk assessment of multi-stage liver fibrosis, namely liver fibrosis and cirrhosis, based on the combined indicators of serum TET3 and FIB-4. Summary of the Invention

[0005] To achieve the above objectives, this application provides a method for constructing a risk assessment model for multi-stage liver fibrosis, comprising the following steps: An observational sample set was constructed, consisting of independent and dependent variables. The independent variables were serum TET3 concentration and FIB-4 index, and the dependent variable was the diagnostic results of multi-stage liver fibrosis. Multi-stage liver fibrosis included F1-F3 and F4 stages. Define a risk assessment model; the risk assessment model is used to assess the risk of developing multi-stage liver fibrosis based on serum TET3 concentration and FIB-4 index; Based on the observed sample set, the risk assessment model is iteratively calculated to determine the optimal stage coefficients for different assessment stages; The optimal stage coefficient is substituted into the probability assessment model to generate a probability assessment formula. The probability assessment formula calculates the probability value of liver fibrosis at each stage based on the serum TET3 concentration and FIB-4 index to be assessed. The probability threshold is then combined to realize the risk assessment of liver fibrosis at each stage.

[0006] The risk assessment model consists of a stage coefficient judgment model and a probability assessment model. The probability assessment model calculates the probability of developing multi-stage liver fibrosis based on the input serum TET3 concentration and FIB-4 index, combined with the stage coefficient; the stage coefficient in the probability assessment model is determined by the stage coefficient judgment model.

[0007] The probability evaluation model is expressed as follows: ,in, Serum TET3 concentration, For FIB-4 index, To assess the risk probability of the liver fibrosis stage; , and The coefficients are the stage coefficients, namely the intercept, TET3 regression coefficient, and FIB-4 regression coefficient.

[0008] Among them, the stage coefficient judgment model is used to determine the optimal stage coefficient for each evaluation stage by observing the sample set; The stage coefficient judgment model is expressed as follows: ,in, For the number of iterations, Let be the parameter estimate vector at the t-th iteration and , For the log-likelihood function in The first derivative vector at that point, For the log-likelihood function in The second derivative matrix at that point, and, ,in, For the observation sample number, For the number of observation samples, For an augmented vector of independent variables that includes the intercept term, The first Serum TET3 concentration and FIB-4 index of each sample For the first The outer product matrix consists of the independent variables of each observation sample.

[0009] Furthermore, the convergence conditions for iterative calculations include: || 、| Or t is greater than the specified number of times; where, For the pre-set convergence tolerance, For the first Log-likelihood function value under the parameters of the next iteration.

[0010] Furthermore, the probability threshold is determined using the Youden index, including: Substitute the stage coefficients into the probability assessment formula to generate the ROC curve; the ROC curve has sensitivity as the vertical axis and specificity as the horizontal axis; where, Youden index = sensitivity + specificity - 1; the probability value corresponding to the maximum value of Youden index is the probability threshold for stage risk assessment.

[0011] Furthermore, the probability evaluation formula is as follows: , , in, The probability of a patient's liver fibrosis progressing to stages F1-F3. The probability that a patient's liver fibrosis will progress to stage F4. This refers to serum TET3 concentration. This is the FIB-4 index.

[0012] Furthermore, the FIB-4 index is calculated as follows: ,in, For platelet count, Aspartate aminotransferase index, For alanine aminotransferase, This is the FIB-4 index.

[0013] According to this invention, the FIB-4 index and serum TET3 concentration can be combined to achieve risk assessment of liver fibrosis and cirrhosis, and the accuracy of the assessment can be improved through continuous application. The risk assessment model generated by this invention can be flexibly embedded into liver disease risk assessment systems. Only indicators such as TET3, AST, ALT, and platelet count obtained from blood tests need to be input to quickly obtain assessment results. This enables self-service screening of liver fibrosis risk anytime and anywhere, lowers the threshold for medical treatment, and is beneficial for early screening and regular follow-up of large populations. Attached Figure Description

[0014] Figure 1 This is a schematic diagram illustrating the construction steps of a risk assessment model for multi-stage liver fibrosis provided according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the ROC curve generated by the probability evaluation model provided in the embodiments of the present invention. Detailed Implementation

[0015] The specific implementation of the present invention will now be described in detail with reference to the accompanying drawings.

[0016] Figure 1 A schematic diagram illustrating the construction steps of a risk assessment model for multi-stage liver fibrosis is provided, as shown in the figure, including the following steps: Step S100: Construct the observation sample set; The observation sample set consists of independent and dependent variables. The independent variables are serum TET3 concentration and FIB-4 index, and the dependent variable is the diagnosis result of multi-stage liver fibrosis.

[0017] This invention provides a specific case: With the patients' consent, clinical data of 212 patients diagnosed with chronic liver disease by liver biopsy were collected, including: patient age, serum TET3 concentration measured by ELISA, AST and ALT levels measured by a fully automated biochemical analyzer, platelet count measured by a fully automated blood cell analyzer, and the diagnostic results of liver fibrosis pathological staging (F1-F3 and F4 stages) determined by liver biopsy as the gold standard. Serum TET3 concentration was used as the independent variable X1, and the FIB-4 index, calculated from patient age, AST, ALT levels, and platelet count, was used as the independent variable X2. The diagnostic results of the pathological stages of liver fibrosis were defined as follows: F1-F3 stage indicates liver fibrosis and F4 stage indicates liver cirrhosis. The diagnostic results of F1-F3 stage and F4 stage were defined as dependent variables Y1 and Y2, respectively.

[0018] Step S110: Define the risk assessment model; the risk assessment model is used to assess the risk of multi-stage liver fibrosis based on serum TET3 concentration and FIB-4 index.

[0019] In this invention, the stages of multi-stage liver fibrosis are defined as F1-F3 and F4, which represent the liver fibrosis stage and the cirrhosis stage, respectively.

[0020] The risk assessment model consists of a stage coefficient judgment model and a probability assessment model. The probability assessment model calculates the probability of multi-stage liver fibrosis based on the input serum TET3 concentration and FIB-4 index, combined with the stage coefficient. The stage coefficient in the probability assessment model is determined by the stage coefficient judgment model.

[0021] 1) The probability assessment model is used to calculate the probability of multi-stage liver fibrosis based on serum TET3 concentration and FIB-4 index combined with stage coefficient; The probability assessment model is based on Logistic regression; Logistic regression is a generalized linear model suitable for regression analysis where the dependent variable is a binary variable. In this invention, the dependent variable is "whether liver fibrosis / cirrhosis has occurred," which is a typical binary category, and the independent variables are serum TET3 concentration and FIB-4 index.

[0022] Logistic regression can be expressed as: ,in, This refers to serum TET3 concentration. For FIB-4 index, For the intercept term, and These are the regression coefficients for each independent variable.

[0023] However, the output of linear regression The value range is (-∞, +∞), which cannot be directly used as the probability of a binary classification event; therefore, the output of the Logistic regression function... Mapping to the interval [0, 1], it can be represented as: ,in, For dependent variable, P For probability; A mathematical transformation is performed to generate a linear regression model between the logarithmic odds ratio and the independent variables, expressed as follows: ,in, As the advantage ratio, This is a Logit transformation.

[0024] Based on the linear regression model, the probability evaluation model is defined as follows: ,in, Serum TET3 concentration, For FIB-4 index, To assess the risk probability of the liver fibrosis stage; , and The coefficients are the stage coefficients, namely the intercept, TET3 regression coefficient, and FIB-4 regression coefficient.

[0025] In practical applications, the optimal values ​​of the stage coefficients differ for different evaluation stages, and the optimal values ​​of the stage coefficients are determined by the stage coefficient judgment model.

[0026] 2) The stage coefficient judgment model is used to determine the optimal stage coefficients for each evaluation stage by observing the sample set; In this invention, the optimal stage coefficients are obtained by fitting and calculating using the maximum likelihood estimation method.

[0027] For the probability evaluation model, let there be a total The nth observation sample, the nth The dependent variable for each sample takes the value of , Indicated as a confirmed case. This indicates that the diagnosis is not yet confirmed. The probability of diagnosis for each sample is Then the first The probability contribution of each sample is hour or hour Unified as: .

[0028] Assuming that all samples are independent, the joint probability of all samples is the product of the probability contributions of each sample, and the likelihood function is expressed as: ,in, The log-likelihood function based on this is expressed as: .

[0029] At this point, the maximum likelihood estimation is used to determine... The parameter vector that reaches its maximum value ,Right now: To solve this, for Taking the partial derivatives of each parameter and setting them to zero, we obtain the likelihood equations, expressed as: .

[0030] Based on this, the stage coefficient judgment model is defined as follows: ,in, For the number of iterations, This is the vector of parameter estimates at the t-th iteration. For the log-likelihood function in The first derivative vector at that point, For the log-likelihood function in The second derivative matrix at that point, and, ,in, For the observation sample number, For the number of observation samples, For an augmented vector of independent variables that includes the intercept term, The first Serum TET3 concentration and FIB-4 index of each sample For the first The outer product matrix consists of the independent variables of each observation sample.

[0031] Step S120: Based on the observed sample set, iteratively calculate the risk assessment model to determine the optimal stage coefficients corresponding to different assessment stages; In this step, the parameter vector in the model is determined by iteratively calculating and solving the stage coefficients. : From the initial guess value Starting from this point, in each iteration, we utilize the first derivative of the objective function. and second derivative The direction and step size of parameter updates are calculated to gradually approach the optimal solution.

[0032] The convergence conditions for iterative calculations include: || 、| Or t is greater than the specified number of times (e.g., 100 times); where, For a pre-defined convergence tolerance (e.g., 10) -6 ), For the first Log-likelihood function value under the parameters of the next iteration.

[0033] The iteration terminates when any of the above convergence conditions are met. That is, the maximum likelihood estimate β̂ of the model parameters, corresponding to This is the optimal stage coefficient of the probability assessment model.

[0034] Since the evaluation stages involved in this invention include F1-F3 and F4, and the corresponding optimal stage coefficients include two sets of probability evaluations for F1-F3 and F4 respectively, probability evaluation formulas for F1-F3 and F4 are generated based on the two sets of optimal stage coefficients respectively.

[0035] In the case provided by this invention, after iterative calculation based on clinical data of patients with chronic liver disease, the following optimal stage coefficients were obtained: Optimal stage coefficients for F1-F3 stages: , , -0.477; The optimal stage coefficient for period F4: , , -2.71.

[0036] Substitute the optimal stage coefficients into the probability assessment formula to generate the ROC curve. Use the AUC value to evaluate the accuracy of the risk assessment model and determine the probability threshold for stage risk assessment.

[0037] Specifically, the ROC curve is plotted with sensitivity as the ordinate and specificity as the abscissa; the closer the AUC value (area under the curve) of the curve is to 1, the higher the prediction accuracy of the model.

[0038] Based on the ROC curve, the probability threshold for stage risk assessment is determined using the Youden index: The Yoden Index is calculated as Sensitivity + Specificity - 1. The maximum value of the Yoden Index is used as the criterion to determine the probability threshold for stage risk assessment.

[0039] In the case provided by the present invention, based on data from 212 patients with chronic liver disease diagnosed by liver biopsy, the optimal stage coefficient determined in step S120 is substituted into the probability evaluation formula to plot the ROC curve.

[0040] Figure 2 Part A presents the ROC curves generated using the probability assessment formula for liver fibrosis (F1-F3 stages); the Youden index (Youden index = sensitivity + specificity - 1) corresponding to different candidate probability thresholds was calculated, yielding the following results:

[0041] As shown in the table above, the Youden index reaches its maximum value (0.703) when the candidate threshold is 0.44. At this point, the sensitivity is 88.7% and the specificity is 81.6%, indicating that the model achieves an optimal balance between sensitivity and specificity. Therefore, 0.44 is determined as the probability threshold for assessing the risk of liver fibrosis (F1-F3 stages). .

[0042] Figure 2 Part B presents the ROC curve generated from the probability assessment formula for liver cirrhosis (F4 stage). The Youden index corresponding to each candidate threshold is calculated using the same method.

[0043] When the candidate threshold is 0.52, the Youden index reaches its maximum value (0.873), at which point the sensitivity is 94.4% and the specificity is 92.9%. Therefore, 0.52 is determined as the probability threshold for risk assessment of cirrhosis (stage F4). .

[0044] The process of determining the aforementioned thresholds ensures an optimal balance between sensitivity and specificity in the evaluation results, resulting in the model having the best diagnostic efficacy in clinical applications. Figure 2 The ROC curves shown have AUC values ​​of 0.943 and 0.990, respectively, indicating that both models have good discriminative ability.

[0045] Step S130: Substitute the optimal stage coefficient into the probability assessment model to generate multiple probability assessment formulas. The multiple probability assessment formulas can calculate the probability value of liver fibrosis at each stage based on the serum TET3 concentration and FIB-4 index to be assessed. Combined with the probability threshold, the risk assessment of liver fibrosis at each stage can be realized.

[0046] At this point, the risk assessment model is complete. In practical applications, as samples are added, the stage coefficients and probability thresholds are continuously optimized, gradually improving the accuracy and effectiveness of the risk assessment.

[0047] In the example provided by this invention, the probability evaluation formula for generating periods F1-F3 and F4 is as follows: , , in, The probability of a patient's liver fibrosis progressing to stage F1-F3 (the development of liver fibrosis). The probability of a patient's liver fibrosis progressing to stage F4 (developing cirrhosis). This refers to serum TET3 concentration. This is the FIB-4 index. Furthermore, and respectively with probability threshold probability threshold Comparison is used to achieve risk assessment of liver fibrosis and cirrhosis.

[0048] The FIB-4 index is typically calculated as follows: ,in, For platelet count, Aspartate aminotransferase index, For alanine aminotransferase, The FIB-4 index is used. Platelet count is obtained using a fully automated hematology analyzer, and the aspartate aminotransferase index and alanine aminotransferase index are obtained using a fully automated biochemical analyzer.

[0049] The risk assessment model constructed by this invention can combine the FIB-4 index and serum TET3 concentration to assess the risk of liver fibrosis and cirrhosis, and its accuracy will be improved through continuous application. The risk assessment model generated by this invention can be flexibly embedded into the liver disease risk assessment system and supports running on WeChat mini programs or mobile apps on mobile terminals (smartphones, tablets). Patients do not need to go to the hospital; they only need to input indicators such as TET3, AST, ALT, and platelet count obtained from blood tests to quickly obtain assessment results. This enables self-service screening of liver fibrosis risk anytime and anywhere, lowers the threshold for medical treatment, and is conducive to early screening and regular follow-up of large-scale populations.

[0050] The above-disclosed embodiments are merely a few specific examples of the present invention. However, the present invention is not limited thereto, and any variations that can be conceived by those skilled in the art should fall within the protection scope of the present invention.

Claims

1. A method for constructing a risk assessment model for multi-stage liver fibrosis, characterized in that, Includes the following steps: An observation sample set was constructed, which consisted of independent and dependent variables. The independent variables were serum TET3 concentration and FIB-4 index, and the dependent variable was the diagnostic results of multi-stage liver fibrosis. The multi-stage liver fibrosis included F1-F3 and F4 stages. Define a risk assessment model; the risk assessment model is used to assess the risk of multi-stage liver fibrosis based on serum TET3 concentration and FIB-4 index; Based on the observed sample set, the risk assessment model is iteratively calculated to determine the optimal stage coefficients corresponding to different assessment stages; The optimal stage coefficient is substituted into the probability assessment model to generate a probability assessment formula. The probability assessment formula calculates the probability value of liver fibrosis at each stage based on the serum TET3 concentration and FIB-4 index to be assessed, and combines the probability threshold to realize the risk assessment of liver fibrosis at each stage.

2. The method for constructing a risk assessment model for multi-stage liver fibrosis according to claim 1, characterized in that, The risk assessment model consists of a stage coefficient judgment model and a probability assessment model; The probability assessment model calculates the probability of multi-stage liver fibrosis based on the input serum TET3 concentration and FIB-4 index, combined with the stage coefficient; the stage coefficient in the probability assessment model is determined by the stage coefficient judgment model.

3. The method for constructing a risk assessment model for multi-stage liver fibrosis according to claim 2, characterized in that, The probability evaluation model is expressed as follows: ,in, Serum TET3 concentration, For FIB-4 index, To assess the risk probability of the liver fibrosis stage; , and The coefficients are the stage coefficients, namely the intercept, TET3 regression coefficient, and FIB-4 regression coefficient.

4. The method for constructing a risk assessment model for multi-stage liver fibrosis according to claim 3, characterized in that, The stage coefficient judgment model is used to determine the optimal stage coefficient for each evaluation stage using the observed sample set; The stage coefficient judgment model is expressed as follows: ,in, For the number of iterations, Let be the parameter estimate vector at the t-th iteration and , For the log-likelihood function in The first derivative vector at that point, For the log-likelihood function in The second derivative matrix at that point, and, ,in, For the observation sample number, For the number of observation samples, For an augmented vector of independent variables that includes the intercept term, The first Serum TET3 concentration and FIB-4 index of each sample For the first The outer product matrix consists of the independent variables of each observation sample.

5. The method for constructing a risk assessment model for multi-stage liver fibrosis according to claim 4, wherein the convergence condition for the iterative calculation includes: || 、| Or t is greater than the specified number of times; where, For the pre-set convergence tolerance, For the first Log-likelihood function value under the parameters of the next iteration.

6. The method for constructing a risk assessment model for multi-stage liver fibrosis according to claim 3, wherein the probability threshold is determined by the Youden index, including: Substituting the stage coefficients into the probability assessment formula generates an ROC curve; the ROC curve has sensitivity as the vertical axis and 1-specificity as the horizontal axis; where, Youden index = sensitivity + specificity - 1; the probability value corresponding to the maximum value of Youden index is the probability threshold for stage risk assessment.

7. The method for constructing a risk assessment model for multi-stage liver fibrosis according to claim 1, characterized in that, The probability evaluation formula is as follows: ， ， in, The probability of a patient's liver fibrosis progressing to stages F1-F3. The probability that a patient's liver fibrosis will progress to stage F4. This refers to serum TET3 concentration. This is the FIB-4 index.

8. The method for constructing a risk assessment model for multi-stage liver fibrosis according to claim 1, characterized in that, The FIB-4 index is calculated as follows: ,in, For platelet count, Aspartate aminotransferase index, For alanine aminotransferase, This is the FIB-4 index.

Images