Pro-c3 combined with ast for grading the degree of liver inflammation in chb patients and a method of constructing the same

Abstract

The application discloses a PRO-C3 combined AST CHB patient noninvasive liver inflammation degree grading system and a construction method thereof. The system comprises a sample collection module, a feature screening module, a model construction module, an efficiency evaluation module and a clinical deployment application module. The sample collection module collects clinical and serological indexes of CHB patients; the feature screening module determines key features independently related to liver inflammation through logistic regression; the model construction module establishes a diagnostic model for predicting liver inflammation of G3 or more based on PRO-C3 and AST; the efficiency evaluation module calculates AUC, decision curve and calibration curve to evaluate clinical net benefit; and the clinical deployment application module constructs an online application system based on a Web framework, supports batch input, real-time operation and structured result output. The application realizes noninvasive and accurate evaluation of liver inflammation of CHB patients, solves the technical problem of difficulty in clinical deployment of complex models, and has the characteristics of low cost, small risk and convenience for promotion.

Description

Technical Field

[0001] This invention relates to the fields of clinical prediction models and machine learning technology, and in particular to a non-invasive liver inflammation severity grading system for CHB patients based on PRO-C3 combined with AST and its construction method. Background Technology

[0002] Chronic hepatitis B (CHB) is a global infectious disease. Long-term infection can lead to liver inflammation, fibrosis, cirrhosis, and even hepatocellular carcinoma. The degree of liver inflammation is a key indicator for assessing disease activity, guiding antiviral treatment decisions, and predicting prognosis. Currently, liver biopsy remains the gold standard for diagnosing liver inflammation, but it has limitations such as invasiveness, sampling errors, high cost, and the risk of postoperative complications, making it difficult to widely use and repeat in clinical practice.

[0003] Serological biomarkers are valuable in liver disease assessment due to their non-invasive, simple, and reproducible advantages. However, the commonly used alanine aminotransferase (ALT) level does not fully reflect the degree of histological inflammation; a significant proportion of CHB patients with significant liver inflammation have normal or only slightly elevated ALT levels. Other non-invasive indices such as APRI and FIB-4 are mainly used to assess liver fibrosis rather than directly quantifying inflammatory activity. Therefore, there is an urgent clinical need for a serological biomarker or diagnostic model that can accurately and non-invasively identify significant liver inflammation in CHB patients.

[0004] PRO-C3 (type III collagen N-terminal propeptide) is a specific cleavage fragment released into the bloodstream during type III collagen synthesis and is a key biomarker reflecting extracellular matrix remodeling and active fibrosis. Recent studies have shown that PRO-C3 is not only closely related to the degree of liver fibrosis but may also respond to liver inflammation-driven dynamic matrix remodeling processes. In non-alcoholic fatty liver disease and metabolic-associated fatty liver disease, PRO-C3-based sequential algorithms have been proven to reliably identify liver inflammation and fibrosis, suggesting potential application value of PRO-C3 in liver inflammation assessment. However, the diagnostic value of PRO-C3 in significant liver inflammation in CHB patients and its potential for combined application with other indicators still lack systematic research.

[0005] Aspartate aminotransferase (AST) is a commonly used enzymatic indicator in clinical practice reflecting hepatocellular damage. It is noteworthy that AST exists in hepatocellular cells in two isoenzymes: c-AST, mainly distributed in the cytoplasm, and m-AST, present in the mitochondria. When hepatocellular damage is mild, the mitochondria remain intact, primarily releasing ALT and a small amount of c-AST; however, when hepatocellular damage is severe or even necrosis occurs, mitochondrial membrane permeability changes, and m-AST is released from the mitochondria into the bloodstream. Therefore, an increase in m-AST is closely related to the depth of inflammatory damage. Furthermore, in chronic viral hepatitis, a significant increase in AST often indicates that chronic hepatitis has entered an active phase. However, current clinical applications of AST mainly focus on detecting total AST levels, failing to fully utilize its subcellular localization characteristics for precise assessment of the depth of inflammatory damage, and there is no systematic approach combining PRO-C3 and AST for the non-invasive diagnosis of significant liver inflammation in CHB patients.

[0006] Even with the development of highly accurate diagnostic models, their widespread application in clinical practice still faces technical obstacles. Existing non-invasive diagnostic models are mostly manually calculated formulas or independent scoring tools. Clinicians need to manually input multiple indicators and perform calculations, a process that is not only inefficient and prone to human error, but also lacks standardized data input and result interpretation procedures, making it difficult to achieve rapid and accurate assessments in busy clinical work. Furthermore, these tools often only output a single score or probability value, lacking structured result presentation and clinical interpretation suggestions, which is detrimental to the standardization and normalization of clinical decision-making. Therefore, how to encapsulate complex diagnostic models into convenient, reliable, and easily deployable clinical tools, achieving a seamless transition from model construction to clinical application, has become a pressing technical problem to be solved in this field. Summary of the Invention

[0007] The purpose of this invention is to provide a non-invasive liver inflammation severity grading system for CHB patients based on PRO-C3 combined with AST and its construction method, so as to achieve non-invasive, accurate and convenient assessment of significant liver inflammation in CHB patients, and solve the technical problem that the model is difficult to apply rapidly in clinical practice through systematic clinical deployment.

[0008] The technical solution to achieve the purpose of this invention is: a non-invasive liver inflammation severity grading system for CHB patients using PRO-C3 combined with AST, comprising a sample collection module, a feature screening module, a model construction module, a performance evaluation module, and a clinical deployment and application module, wherein:

[0009] The sample collection module collects clinical and laboratory information from medical records of patients with chronic hepatitis B (CHB) who undergo liver biopsy.

[0010] The feature screening module uses a logistic regression algorithm to screen key clinical features that are independently associated with liver inflammation G≥3; where G represents the Scheuer system score level, and G≥3 indicates that the patient's liver biopsy shows significant liver inflammation.

[0011] The model building module uses a logistic regression algorithm to incorporate two clinical features, PRO-C3 and AST, determined by the feature screening module in the training queue, to establish a predictive model for liver inflammation G≥3; PRO-C3 represents type III collagen amino-terminal propeptide, and AST represents aspartate aminotransferase.

[0012] The efficacy evaluation module calculates the area under the receiver operating characteristic curve for the predictive model to diagnose liver inflammation with G≥3, and evaluates the clinical net benefit through decision curve analysis and calibration curve analysis.

[0013] The clinical deployment and application module, as the deployment and application unit of the prediction model, is used to encapsulate the prediction model into a clinical decision support tool that can be remotely invoked. The internal units of this tool are deeply integrated with the prediction model output by the model building module to achieve automated processing of clinical indicators and real-time diagnostic output.

[0014] Furthermore, the sample acquisition module is specifically used for:

[0015] We collected data from multiple hospitals on newly diagnosed CHB patients who underwent liver biopsy. CHB refers to patients who have been positive for hepatitis B surface antigen (HBsAg) for at least 6 months.

[0016] Clinical and laboratory information related to patients was retrieved from the medical records of participating hospitals, and serum PRO-C3 levels were detected using an enzyme-linked immunosorbent assay (ELISA).

[0017] Furthermore, the feature filtering module is specifically used for:

[0018] Using logistic regression, univariate and multivariate logistic regression analyses were performed on the following 20 collected clinical features to screen out key clinical features independently associated with liver inflammation G≥3:

[0019] Age, sex, body mass index, white blood cell count, hemoglobin, platelet count, total bilirubin, albumin, alanine aminotransferase, aspartate aminotransferase, AST / ALT ratio, alkaline phosphatase, gamma-glutamyl transferase, hepatitis B e antigen status, hepatitis B virus deoxyribonucleic acid, type III collagen amino-terminal propeptide, FIB-4 index, aspartate aminotransferase to platelet ratio, liver stiffness test, and controlled attenuation parameters.

[0020] Furthermore, the model building module is specifically used for:

[0021] Based on the results of multivariate logistic regression analysis, clinical features independently associated with liver inflammation G≥3 were screened, including PRO-C3, AST, and total bilirubin.

[0022] PRO-C3 and AST were selected as model predictors to establish a logistic regression model. This logistic regression model was then used as a predictive model for diagnosing liver inflammation G≥3.

[0023] Furthermore, the formula for the logistic regression model established in the model building module is as follows:

[0024]

[0025] in, Indicates the probability of an event occurring. of Transformation, The probability of a patient developing significant liver inflammation (G≥3). This indicates the PRO-C3 index value obtained through enzyme-linked immunosorbent assay (ELISA). This represents the AST index value obtained through blood biochemistry testing; -4.641196 is the intercept of the logistic regression model; and 0.051447 is... The regression coefficient is 0.032088. The regression coefficients are obtained by fitting the data from the training queue.

[0026] Furthermore, the performance evaluation module is specifically used for:

[0027] Calculate the area under the receiver operating characteristic curve for the predictive model to diagnose liver inflammation with G≥3;

[0028] The clinical net benefit of the prediction model under different threshold probabilities was evaluated through decision curve analysis and compared with traditional serological markers.

[0029] The consistency between the predicted probability of the prediction model and the actual observed probability was evaluated by calibration curve analysis and Brier score.

[0030] Furthermore, the calculation results of the efficacy evaluation module show that the area under the receiver operating characteristic curve for the predictive model to diagnose liver inflammation G≥3 is 0.873, the sensitivity is 84.3%, the specificity is 77.2%, and the negative predictive value is 94.7%.

[0031] Furthermore, the clinical deployment application module specifically includes:

[0032] The data receiving unit is used to receive PRO-C3 and AST clinical indicator data that are entered in a single instance or imported in batches. It supports manual entry for a single case and batch uploading of Excel files containing indicators from up to 20 patients.

[0033] A data preprocessing unit, connected to the data receiving unit, is used to process the received PRO-C3 and AST clinical indicator data for missing value processing, outlier detection, and standardization.

[0034] The model computation unit has the built-in logistic regression model and is directly connected to the data preprocessing unit. It is used to substitute the preprocessed standardized data into the logistic regression model for real-time concurrent computation and output the risk probability of significant liver inflammation G≥3 for each patient.

[0035] The result output unit, connected to the model operation unit, is used to present the risk probability in the form of visual charts and structured documents. It supports generating diagnostic reports containing patient information, model input values, predicted probabilities, and clinical suggestions based on preset thresholds, and supports exporting the results to Excel or PDF documents.

[0036] Through the collaborative work of the aforementioned units, the clinical deployment application module encapsulates the prediction model into a clinical tool.

[0037] Furthermore, the system also includes a processor and a memory, the memory storing a computer program, and the processor executing the computer program to implement the functions of the sample acquisition module, feature screening module, model building module, efficacy evaluation module, and clinical deployment application module.

[0038] A method for constructing a non-invasive liver inflammation severity grading system for CHB patients using PRO-C3 combined with AST, as described above, comprising the following steps:

[0039] Step 1: Construct a sample collection module to collect clinical and laboratory information from medical records of CHB patients undergoing liver biopsy;

[0040] Step 2: Construct a feature screening module to screen key clinical features that are independently associated with liver inflammation G≥3 using a logistic regression algorithm;

[0041] Step 3: Build the model building module. Based on the PRO-C3 and AST features selected in Step 2, use the logistic regression algorithm to build a predictive model for diagnosing liver inflammation with G≥3.

[0042] Step 4: Construct a performance evaluation module, calculate the area under the receiver operating characteristic curve for the predictive model to diagnose liver inflammation with G≥3, and evaluate the clinical net benefit through decision curve analysis and calibration curve analysis.

[0043] Step 5: Construct a clinical deployment application module, which encapsulates the prediction model established in Step 3 into a remotely invoked clinical decision support tool. The clinical deployment application module includes a data receiving unit, a model operation unit, and a result output unit. The model is deployed and applied in the clinical environment through the collaborative work of each unit.

[0044] Compared with the prior art, the significant advantages of this invention are:

[0045] (1) This invention is the first to demonstrate through multivariate logistic regression that PRO-C3 is an independent predictor of significant liver inflammation (G≥3) in CHB patients, and a non-invasive diagnostic model of PRO-C3 combined with AST was constructed, filling the gap in accurate non-invasive assessment in this field;

[0046] (2) The PRO-C3 combined with AST model constructed in this invention has excellent diagnostic efficacy (AUC=0.873) and extremely high negative predictive value (94.7%). It can be used as an ideal "exclusion" tool to effectively reduce unnecessary liver biopsies.

[0047] (3) Through decision curve analysis and calibration curve verification, the model of the present invention has significant clinical net benefit and good predictive accuracy, which is superior to single PRO-C3, AST and other traditional serological markers;

[0048] (4) The present invention innovatively designs a clinical deployment application module. Through the collaborative work of data receiving, model calculation and result output units, the complex diagnostic model is encapsulated into a convenient clinical tool, which solves the technical problem that the diagnostic model is difficult to deploy quickly and apply accurately in clinical practice. The input verification mechanism built into this module can automatically identify abnormal values ​​and ensure the reliability of diagnostic results. The batch processing function greatly improves the efficiency of clinical work, making the technical solution of the present invention have complete clinical translation value. Attached Figure Description

[0049] Figure 1 This is a structural diagram of the non-invasive liver inflammation severity grading system for CHB patients using PRO-C3 combined with AST.

[0050] Figure 2 This is a comparison of receiver operating characteristic curves (ROCs) for diagnosing significant liver inflammation using PRO-C3 combined with the AST model and other serological markers.

[0051] Figure 3 This is a comparison chart of decision curve analysis of the PRO-C3 combined with AST model and other serological markers.

[0052] Figure 4 This is a calibration curve of the PRO-C3 combined with AST model and other serological markers.

[0053] Figure 5 This is a forest plot showing the diagnostic efficacy of the PRO-C3 combined with AST model in different subgroups (fibrosis stage, virological status, HBV DNA level, ALT level). Detailed Implementation

[0054] This invention proposes a machine learning model based on PRO-C3 and AST for non-invasive diagnosis of significant liver inflammation (G≥3) in CHB patients, and solves the technical problem of the model's difficulty in rapid application in clinical practice through an innovative clinical deployment application module.

[0055] Combination Figure 1 This invention discloses a non-invasive liver inflammation severity grading system for CHB patients using PRO-C3 combined with AST, comprising a sample collection module, a feature screening module, a model construction module, a performance evaluation module, and a clinical deployment and application module, wherein:

[0056] The sample collection module collects clinical and laboratory information from medical records of CHB patients undergoing liver biopsy; CHB refers to chronic hepatitis B.

[0057] The feature screening module uses a logistic regression algorithm to screen key clinical features that are independently associated with liver inflammation (G≥3); G represents the grade, and G≥3 means that the patient's liver biopsy Scheuer system score is not less than grade 3;

[0058] The model building module uses a machine learning algorithm to incorporate two clinical features, PRO-C3 and AST, from the training cohort to establish a predictive model for diagnosing liver inflammation G≥3; PRO-C3 represents type III collagen amino-terminal propeptide, and AST represents aspartate aminotransferase.

[0059] The efficacy evaluation module calculates the optimal area under the receiver operating characteristic curve for the model to diagnose liver inflammation, and evaluates the clinical net benefit through decision curve analysis and calibration curve analysis.

[0060] The clinical deployment and application module, as the deployment and application unit of the diagnostic model, is used to encapsulate the prediction model into a remotely invoked clinical decision support tool. Its internal units are deeply integrated with the prediction model output by the model building module to achieve automated processing of clinical indicators and real-time diagnostic output.

[0061] As a specific example, the sample acquisition module is as follows:

[0062] We collected data from multiple hospitals on newly diagnosed CHB patients who underwent liver biopsy. CHB refers to patients who have been positive for hepatitis B surface antigen (HBsAg) for at least 6 months.

[0063] Clinical and laboratory information related to patients was retrieved from the medical records of participating hospitals, and serum PRO-C3 levels were detected using ELISA (enzyme-linked immunosorbent assay).

[0064] As a specific example, the feature filtering module is specifically used for:

[0065] The logistic regression machine learning algorithm was used to incorporate 20 collected features: age, gender, body mass index, white blood cell count, hemoglobin, platelet count, total bilirubin, albumin, ALT (alanine aminotransferase), AST, AST / ALT, alkaline phosphatase, gamma-glutamyl transferase, hepatitis B e antigen status, hepatitis B virus deoxyribonucleic acid, PRO-C3, FIB-4 index, aspartate aminotransferase to platelet ratio, liver stiffness test, and controlled decay parameters.

[0066] Univariate and multivariate logistic regression algorithms were used to screen key clinical features independently associated with liver inflammation (G≥3). Results showed that PRO-C3 (OR=1.114, 95%CI: 1.036-1.241, P=0.014), AST (OR=1.127, 95%CI: 1.012-1.298, P=0.042), and total bilirubin (OR=1.149, 95%CI: 1.046-1.302, P=0.010) were significant independent predictors of liver inflammation. Based on model simplicity and clinical applicability, PRO-C3 and AST were selected as the final modeling features.

[0067] As a specific example, the model building module is specifically used for:

[0068] Based on the selected PRO-C3 and AST features, a logistic regression model is constructed. The model's formula is expressed as follows:

[0069]

[0070] in, Indicates the probability of an event occurring. of Transformation, The probability of a patient developing significant liver inflammation (G≥3). This indicates the PRO-C3 index value obtained through enzyme-linked immunosorbent assay (ELISA). This represents the AST index value obtained through blood biochemistry testing; -4.641196 is the intercept of the logistic regression model; and 0.051447 is... The regression coefficient is 0.032088. The regression coefficients are obtained by fitting the data from the training queue.

[0071] As a specific example, the performance evaluation module is specifically used for:

[0072] The diagnostic efficacy of the model was evaluated using receiver operating characteristic (ROC) curve analysis. In the training cohort, the AUC of the PRO-C3 combined with the AST model for diagnosing significant liver inflammation was 0.873 (95% CI: 0.829-0.917), with a sensitivity of 84.3%, a specificity of 77.2%, and a negative predictive value as high as 94.7%.

[0073] The clinical net benefit of the model was evaluated through decision curve analysis. The results showed that the PRO-C3 combined with AST model had a higher net benefit than PRO-C3, AST alone or other traditional serological markers (such as ALT, GGT, PLT, etc.) over a wide range of threshold probabilities.

[0074] The model's calibration was assessed using calibration curves and Brier scores. The calibration curve of the PRO-C3 combined with AST model was close to the ideal diagonal, and the Brier score was 0.105, lower than other single markers, indicating a high degree of consistency between its predicted probability and the actual probability of occurrence.

[0075] As a specific example, the clinical deployment application module specifically includes:

[0076] The data receiving unit is used to receive PRO-C3 and AST clinical indicator data, either input individually or imported in batches. This unit provides two input modes: single-case manual entry mode, allowing doctors to quickly input data for a single patient during consultations; and batch upload mode, supporting the simultaneous upload of data from up to 20 patients in Excel file format, meeting the needs of clinical research or batch screening.

[0077] The model computation unit incorporates the aforementioned trained and validated logistic regression diagnostic model and directly interfaces with the data preprocessing unit. This unit receives preprocessed, standardized data, invokes the encapsulated model for real-time concurrent computation, and quickly outputs the probability (P-value) of significant liver inflammation risk for each patient. For batch-uploaded data, this unit supports parallel computation, significantly improving processing efficiency.

[0078] The results output unit, connected to the model computation unit, is used to present the computation results in multiple formats. This unit can generate visual risk probability charts to intuitively display the patient's inflammation risk level; simultaneously, it can generate structured diagnostic reports, including basic patient information, model input values, predicted probabilities, and risk grading and clinical recommendations based on preset clinical thresholds (e.g., with 0.5 as the cutoff). The reports can be exported as Excel or PDF documents for easy archiving, printing, or communication with patients by doctors.

[0079] Through the close collaboration of the aforementioned units, the clinical deployment application module encapsulates complex diagnostic models into a practical tool that is easy to operate, reliable in results, and conforms to clinical workflows, solving the "last mile" technical challenge of bringing high-precision diagnostic models from the laboratory to clinical applications.

[0080] As a concrete example, the entire system is designed in accordance with medical data security standards, uses encrypted protocols for data transmission, and has access control to ensure patient privacy and data security.

[0081] This invention also provides a method for constructing a non-invasive liver inflammation severity grading system for CHB patients using PRO-C3 combined with AST, the method comprising the following steps:

[0082] Step 1: Construct a sample collection module. This module collects clinical and laboratory information from medical records of CHB patients undergoing liver biopsy.

[0083] Step 2: Construct a feature screening module to screen key clinical features strongly correlated with liver inflammation using a logistic regression algorithm;

[0084] Step 3: Build the model building module. Based on the PRO-C3 and AST features selected in Step 2, use the logistic regression algorithm to build a predictive model for diagnosing liver inflammation.

[0085] Step 4: Construct a performance evaluation module, calculate the optimal area under the receiver operating characteristic curve for liver inflammation diagnosed by the diagnostic model, and evaluate the clinical net benefit through decision curve analysis and calibration curve analysis.

[0086] Step 5: Construct a clinical deployment application module, which encapsulates the prediction model established in Step 3 into a remotely invoked clinical decision support tool. The clinical deployment application module includes a data receiving unit, a model operation unit, and a result output unit. Through the collaborative work of each unit, the model can be quickly deployed and accurately applied in the clinical environment.

[0087] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the steps in the method for constructing the PRO-C3 combined with AST non-invasive liver inflammation severity grading system for CHB patients.

[0088] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. The present invention also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps in the method for constructing the PRO-C3 combined with AST non-invasive liver inflammation severity grading system for CHB patients.

[0089] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0090] Example

[0091] This embodiment provides a non-invasive liver inflammation severity grading system for CHB patients using PRO-C3 combined with AST. The system is constructed by including the following components:

[0092] 1. Sample collection

[0093] A retrospective study was conducted on treatment-naïve CHB patients (HBsAg positive for at least 6 months) who underwent liver biopsy at multiple hospitals in China. In this example, the training dataset included 324 patients, of whom 70 (21.6%) were pathologically diagnosed with significant liver inflammation (Scheuer grade G3-4).

[0094] 2. Data Preparation

[0095] Clinical and laboratory information related to patients was retrieved from the medical records of participating hospitals. A total of 20 clinical characteristics were collected: age, sex, body mass index, white blood cell count, hemoglobin, platelet count, total bilirubin, albumin, ALT, AST, AST / ALT ratio, alkaline phosphatase, gamma-glutamyl transferase, hepatitis B e antigen status, hepatitis B virus DNA, PRO-C3, FIB-4 index, APRI, liver stiffness measurement, and controlled attenuation parameters. PRO-C3 was detected using an ELISA method.

[0096] 3. Feature selection and model building

[0097] (1) Feature selection was performed using a logistic regression model. First, univariate logistic regression was performed to screen for potential predictors associated with significant liver inflammation (G≥3) (P<0.05). These factors were then incorporated into a multivariate logistic regression model. Multivariate analysis showed that PRO-C3 (OR=1.114, 95%CI: 1.036-1.241, P=0.014), AST (OR=1.127, 95%CI: 1.012-1.298, P=0.042), and total bilirubin (OR=1.149, 95%CI: 1.046-1.302, P=0.010) were independent predictors of significant liver inflammation. The feature selection results are shown in Table 1.

[0098] Table 1. Multivariate logistic regression analysis screened features that were independently associated with significant liver inflammation.

[0099]

[0100] (2) A predictive model for diagnosing significant liver inflammation (G≥3) was constructed based on PRO-C3 and AST. Considering the simplicity and clinical applicability of the model, PRO-C3 and AST were selected as the final modeling features to construct a logistic regression model. The model formula is as follows:

[0101] logit(P) = -4.641196 + 0.051447 × PROC3 + 0.032088 × AST;

[0102] Where P is the probability of a patient developing significant liver inflammation (G≥3), -4.641196 is the intercept, and 0.051447 and 0.032088 are the regression coefficients of PRO-C3 and AST, respectively, obtained by fitting the training cohort data.

[0103] 4. Evaluation of model performance

[0104] After successfully creating the PRO-C3 joint AST diagnostic model, its diagnostic efficacy in the training cohort was evaluated.

[0105] Receiver operating characteristic (ROC) curve analysis showed that the AUC of PRO-C3 combined with the AST model for diagnosing significant liver inflammation was 0.873 (95% CI: 0.829-0.917), significantly higher than that of PRO-C3 alone (AUC=0.806), AST (AUC=0.808), ALT (AUC=0.712), and other traditional indicators. Figure 2 The optimal cutoff value for the model corresponds to a sensitivity of 84.3%, a specificity of 77.2%, and a negative predictive value as high as 94.7%.

[0106] Decision curve analysis showed that the PRO-C3 combined with AST model had higher clinical net benefit than single markers and other traditional indicators within a wide range of threshold probabilities, indicating its superior clinical utility value ( Figure 3 ).

[0107] Calibration curve analysis showed that there was good consistency between the predicted probability and the actual observed probability of the PRO-C3 combined with AST model, with a Brier score of 0.105, indicating excellent model calibration ( Figure 4 ).

[0108] 5. Subgroup analysis to verify the robustness of the model

[0109] To verify the robustness of the model, the diagnostic efficacy of the PRO-C3 combined with AST model was evaluated in different clinical subgroups. The results showed that the AUC values of the model remained stable regardless of the fibrosis stage (≥S2 or <S2), HBeAg status (positive or negative), HBV DNA level (high or low), and ALT level (normal or elevated) of the patients, indicating good generalization ability ( Figure 5 ). Especially in patients with normal ALT, the model still maintained high diagnostic accuracy, which was of great significance for identifying "gray zone" patients.

[0110] 6. Development of the clinical deployment application module

[0111] Based on the previously constructed and verified PRO-C3 combined with AST diagnostic model, a clinical deployment application module was developed and encapsulated as a decision support tool that can be conveniently used in the clinical environment.

[0112] This module adopts a B / S architecture, and the server side is deployed on the Linux CentOS 8 operating system, using Nginx 1.20 and Gunicorn 20.1 as the running support environment. The module internally contains four core units:

[0113] Data receiving unit: Develop a Web front-end interface, provide a singleton data entry form, and support batch uploading of Excel files (≤20 cases) containing PRO-C3 and AST indicators.

[0114] Model operation unit: Use machine learning libraries in Python (such as scikit-learn) to serialize and save the trained logistic regression model (including intercept and coefficients , ) and encapsulate it as an API interface. This unit receives the data transmitted by the preprocessing unit, calls the model for real-time probability calculation, and returns the risk probability value.

[0115] Results Output Unit: Visualizes the probability values ​​returned by the model calculation unit, displaying them as progress bars or percentages on the front-end page. It also generates a structured report containing patient ID, PRO-C3, AST, predicted probability, and risk level (e.g., probability <0.2 for low risk, 0.2-0.6 for medium risk, and >0.6 for high risk). It supports one-click export of current results or batch results to Excel or PDF files.

[0116] Through the collaborative work of the aforementioned units, this clinical deployment application module transforms complex statistical models into a simple, fast-responding, and intuitive clinical tool, effectively solving the technical challenges of model transformation from research to application.

[0117] In summary, this invention employs logistic regression combined with PRO-C3 and AST indices to construct a non-invasive diagnostic system for significant liver inflammation in CHB patients. This system exhibits excellent diagnostic efficacy and high negative predictive value, effectively identifying patients requiring intervention and reducing unnecessary liver biopsies. Through an innovative clinical deployment module, this invention encapsulates a complex diagnostic model into a convenient clinical tool, solving the technical problem of the difficulty in rapidly applying models in clinical practice. It provides a low-cost, non-invasive, convenient, and easily scalable overall technical solution for the precise management of CHB patients.

[0118] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A non-invasive liver inflammation severity grading system for CHB patients using PRO-C3 combined with AST, characterized in that, It includes a sample collection module, a feature selection module, a model building module, a performance evaluation module, and a clinical deployment and application module, among which: The sample collection module collects clinical and laboratory information from medical records of patients with chronic hepatitis B (CHB) who undergo liver biopsy. The feature screening module uses a logistic regression algorithm to screen key clinical features that are independently associated with liver inflammation G≥3; where G represents the Scheuer system score level, and G≥3 indicates that the patient's liver biopsy shows significant liver inflammation. The model building module uses a logistic regression algorithm to incorporate two clinical features, PRO-C3 and AST, determined by the feature screening module in the training queue, to establish a predictive model for liver inflammation G≥3; PRO-C3 represents type III collagen amino-terminal propeptide, and AST represents aspartate aminotransferase. The efficacy evaluation module calculates the area under the receiver operating characteristic curve for the predictive model to diagnose liver inflammation with G≥3, and evaluates the clinical net benefit through decision curve analysis and calibration curve analysis. The clinical deployment and application module, as the deployment and application unit of the prediction model, is used to encapsulate the prediction model into a clinical decision support tool that can be remotely invoked. The internal units of this tool are deeply integrated with the prediction model output by the model building module to achieve automated processing of clinical indicators and real-time diagnostic output.

2. The PRO-C3 combined with AST non-invasive liver inflammation severity grading system for CHB patients according to claim 1, characterized in that, The sample acquisition module is specifically used for: We collected data from multiple hospitals on newly diagnosed CHB patients who underwent liver biopsy. CHB refers to patients who have been positive for hepatitis B surface antigen (HBsAg) for at least 6 months. Clinical and laboratory information related to patients was retrieved from the medical records of participating hospitals, and serum PRO-C3 levels were detected using an enzyme-linked immunosorbent assay (ELISA).

3. The PRO-C3 combined with AST non-invasive liver inflammation severity grading system for CHB patients according to claim 2, characterized in that, The feature filtering module is specifically used for: Using logistic regression, univariate and multivariate logistic regression analyses were performed on the following 20 collected clinical features to screen out key clinical features independently associated with liver inflammation G≥3: age The following parameters were measured: gender, body mass index, white blood cell count, hemoglobin, platelet count, total bilirubin, albumin, alanine aminotransferase, aspartate aminotransferase, AST / ALT ratio, alkaline phosphatase, gamma-glutamyl transferase, hepatitis B e antigen status, hepatitis B virus deoxyribonucleic acid, type III collagen amino-terminal propeptide, FIB-4 index, aspartate aminotransferase to platelet ratio, liver stiffness test, and controlled attenuation parameters.

4. The PRO-C3 combined with AST non-invasive liver inflammation severity grading system for CHB patients according to claim 3, characterized in that, The model building module is specifically used for: Based on the results of multivariate logistic regression analysis, clinical features independently associated with liver inflammation G≥3 were screened, including PRO-C3, AST, and total bilirubin. PRO-C3 and AST were selected as model predictors to establish a logistic regression model. This logistic regression model was then used as a predictive model for diagnosing liver inflammation G≥3.

5. The PRO-C3 combined with AST non-invasive liver inflammation severity grading system for CHB patients according to claim 4, characterized in that, The formula for the logistic regression model built in the model building module is as follows: in, Indicates the probability of an event occurring. of Transformation, The probability of a patient developing significant liver inflammation (G≥3). This indicates the PRO-C3 index value obtained through enzyme-linked immunosorbent assay (ELISA). This represents the AST index value obtained through blood biochemistry testing; -4.641196 is the intercept of the logistic regression model; and 0.051447 is... The regression coefficient is 0.032088. The regression coefficients are obtained by fitting the data from the training queue.

6. The PRO-C3 combined with AST non-invasive liver inflammation severity grading system for CHB patients according to claim 5, characterized in that, The performance evaluation module is specifically used for: Calculate the area under the receiver operating characteristic curve for the predictive model to diagnose liver inflammation with G≥3; The clinical net benefit of the prediction model under different threshold probabilities was evaluated through decision curve analysis and compared with traditional serological markers. The consistency between the predicted probability of the prediction model and the actual observed probability was evaluated by calibration curve analysis and Brier score.

7. The PRO-C3 combined with AST non-invasive liver inflammation severity grading system for CHB patients according to claim 6, characterized in that, The calculation results of the efficacy evaluation module show that the area under the receiver operating characteristic curve for the predictive model in diagnosing liver inflammation with G≥3 is 0.873, the sensitivity is 84.3%, the specificity is 77.2%, and the negative predictive value is 94.7%.

8. The PRO-C3 combined with AST non-invasive liver inflammation severity grading system for CHB patients according to claim 5, characterized in that, The clinical deployment application module specifically includes: The data receiving unit is used to receive PRO-C3 and AST clinical indicator data that are entered in a single instance or imported in batches. It supports manual entry for a single case and batch uploading of Excel files containing indicators from up to 20 patients. A data preprocessing unit, connected to the data receiving unit, is used to process the received PRO-C3 and AST clinical indicator data for missing value processing, outlier detection, and standardization. The model computation unit has the built-in logistic regression model and is directly connected to the data preprocessing unit. It is used to substitute the preprocessed standardized data into the logistic regression model for real-time concurrent computation and output the risk probability of significant liver inflammation G≥3 for each patient. The result output unit, connected to the model operation unit, is used to present the risk probability in the form of visual charts and structured documents. It supports generating diagnostic reports containing patient information, model input values, predicted probabilities, and clinical suggestions based on preset thresholds, and supports exporting the results to Excel or PDF documents. Through the collaborative work of the aforementioned units, the clinical deployment application module encapsulates the prediction model into a clinical tool.

9. The PRO-C3 combined with AST non-invasive liver inflammation severity grading system for CHB patients according to claim 1, characterized in that, The system also includes a processor and a memory, the memory storing a computer program, which the processor executes to implement the functions of the sample acquisition module, feature screening module, model building module, performance evaluation module, and clinical deployment application module.

10. A method for constructing a non-invasive liver inflammation severity grading system for CHB patients using PRO-C3 combined with AST as described in any one of claims 1 to 9, characterized in that, The method includes the following steps: Step 1: Construct a sample collection module to collect clinical and laboratory information from medical records of CHB patients undergoing liver biopsy; Step 2: Construct a feature screening module to screen key clinical features that are independently associated with liver inflammation G≥3 using a logistic regression algorithm; Step 3: Build the model building module. Based on the PRO-C3 and AST features selected in Step 2, use the logistic regression algorithm to build a predictive model for diagnosing liver inflammation with G≥3. Step 4: Construct a performance evaluation module, calculate the area under the receiver operating characteristic curve for the predictive model to diagnose liver inflammation with G≥3, and evaluate the clinical net benefit through decision curve analysis and calibration curve analysis. Step 5: Construct a clinical deployment application module, which encapsulates the prediction model established in Step 3 into a remotely invoked clinical decision support tool. The clinical deployment application module includes a data receiving unit, a model operation unit, and a result output unit. The model is deployed and applied in the clinical environment through the collaborative work of each unit.

Images