A risk prediction system for hepatocellular carcinoma in patients with chronic hepatitis B

Abstract

This invention discloses a risk prediction system for hepatocellular carcinoma in patients with chronic hepatitis B, comprising a data acquisition unit, a parameter generation unit, an imbalance correction unit, a scoring unit, a probability conversion unit, and a risk output unit. The system acquires the age, sex, quantitative value of hepatitis B surface antigen, albumin value, aspartate aminotransferase value, and platelet count of patients who have achieved sustained virological response after nucleoside or nucleotide analog treatment. It generates hepatitis B surface antigen risk parameters and APRI risk parameters, and when the risk levels of these two parameters are inconsistent, it generates a virus-fibrosis imbalance correction coefficient to adjust the baseline risk score, obtaining a comprehensive risk score. This score then generates the probability of hepatocellular carcinoma occurrence and risk stratification results at three, five, and seven years. This invention can improve the targeting and accuracy of hepatocellular carcinoma risk prediction in treated chronic hepatitis B patients.

Description

Technical Field

[0001] This invention relates to the field of medical data processing technology, and in particular to a system for predicting the risk of hepatocellular carcinoma in patients with chronic hepatitis B. Background Technology

[0002] Chronic hepatitis B is a significant associated disease with hepatocellular carcinoma (HCC). Clinically, antiviral therapy is typically used to suppress hepatitis B virus replication, and patients are followed up and managed long-term based on factors such as age, sex, liver function indicators, platelet count, liver fibrosis-related indicators, and imaging findings. Current HCC risk assessment methods are usually based on clinical scoring scales, statistical scoring systems, or electronic medical record data processing systems. These methods use basic demographic information, laboratory test results, and liver disease progression-related indicators as inputs to obtain risk scores or risk stratification results, which help determine follow-up frequency and screening intensity.

[0003] The main problem with existing technologies is that, for chronic hepatitis B patients who have received nucleoside or nucleotide analogue therapy and achieved sustained virological response, relying solely on routine liver function indicators, platelet counts, or liver fibrosis-related scores is insufficient to fully reflect the difference between residual virological risk and liver fibrosis risk. Furthermore, while quantitative hepatitis B surface antigen (HBsAg) values ​​can reflect certain virological characteristics, they are generally not effectively coordinated with fibrosis-related indicators such as APRI in current risk assessments. When the virological risk level and the fibrosis risk level are inconsistent, it can easily lead to an overestimation or underestimation of risk scores, thus affecting the accuracy of hepatocellular carcinoma probability prediction and risk stratification results.

[0004] Therefore, it is necessary to propose a technical solution for predicting the risk of hepatocellular carcinoma in patients with chronic hepatitis B who have undergone treatment and achieved sustained virological response. Summary of the Invention

[0005] This application provides a risk prediction system for hepatocellular carcinoma in patients with chronic hepatitis B, in order to improve the relevance and accuracy of risk prediction for hepatocellular carcinoma in treated patients with chronic hepatitis B.

[0006] This application provides a system for predicting the risk of hepatocellular carcinoma in patients with chronic hepatitis B, including: The data acquisition unit is used to acquire the age, sex, quantitative value of hepatitis B surface antigen, albumin value, aspartate aminotransferase value, upper limit of normal value of aspartate aminotransferase, and platelet count of chronic hepatitis B patients who have achieved sustained virological response after nucleoside or nucleotide analog treatment, forming a target patient dataset. The parameter generation unit is used to generate hepatitis B surface antigen risk parameters based on the quantitative value of hepatitis B surface antigen in the target patient dataset, and to calculate APRI value and APRI risk parameters based on aspartate aminotransferase value, upper limit of normal aspartate aminotransferase value and platelet count. The imbalance correction unit is used to compare the risk level difference between the hepatitis B surface antigen risk parameter and the APRI risk parameter, and to generate a virus-fibrosis imbalance correction coefficient when the risk level corresponding to the hepatitis B surface antigen risk parameter is inconsistent with the risk level corresponding to the APRI risk parameter. The scoring unit is used to generate a baseline risk score based on age, gender, albumin level, hepatitis B surface antigen risk parameter, and APRI risk parameter, and then corrects the baseline risk score based on the virus-fibrosis imbalance correction coefficient to obtain a comprehensive risk score. The probability conversion unit is used to generate multiple time window probabilities based on the comprehensive risk score, including the probability of occurrence of hepatocellular carcinoma in three years, five years, and seven years. The risk output unit is used to generate risk stratification results based on the occurrence probability of multiple time windows and preset stratification thresholds.

[0007] This application has the following beneficial technical effects: First, by limiting the patient population to those who have received antiviral treatment and achieved sustained virological response, the target group for risk prediction is more clearly defined, improving the applicability of the results. Second, the quantitative value of hepatitis B surface antigen and the APRI value are combined into risk parameters, taking into account both the risk of virological residue and the risk of liver fibrosis. Third, when the risk levels of the two are inconsistent, a virus-fibrosis imbalance correction coefficient is introduced to reduce the scoring bias caused by a single indicator. Fourth, the output of the probability of occurrence and risk stratification results for three-year, five-year, and seven-year time windows is conducive to the development of differentiated follow-up and screening programs. Attached Figure Description

[0008] Figure 1 This is a schematic diagram of a risk prediction system for hepatocellular carcinoma in patients with chronic hepatitis B, provided in the first embodiment of this application. Detailed Implementation

[0009] Many specific details are set forth in the following description to provide a full understanding of this application. However, this application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this application; therefore, this application is not limited to the specific embodiments disclosed below.

[0010] The first embodiment of this application provides a system for predicting the risk of hepatocellular carcinoma in patients with chronic hepatitis B. Please refer to [link / reference]. Figure 1 This figure is a schematic diagram of the first embodiment of this application. The following is in conjunction with... Figure 1The first embodiment of this application provides a risk prediction system for hepatocellular carcinoma in patients with chronic hepatitis B.

[0011] The hepatocellular carcinoma risk prediction system for patients with chronic hepatitis B includes a data acquisition unit 101, a parameter generation unit 102, an imbalance correction unit 103, a scoring unit 104, a probability conversion unit 105, and a risk output unit.

[0012] The acquisition unit 101 is used to acquire the age, sex, quantitative value of hepatitis B surface antigen, albumin value, aspartate aminotransferase value, upper limit of normal value of aspartate aminotransferase, and platelet count of chronic hepatitis B patients who have achieved sustained virological response after treatment with nucleoside or nucleotide analogs, and to form a target patient dataset.

[0013] In this embodiment, the data acquisition unit 101 can be located in a server, hospital information system terminal, follow-up management platform, or independent computing device to acquire and organize basic information and test results of patients with chronic hepatitis B before risk prediction begins. The chronic hepatitis B patient refers to a patient who has a history of or current hepatitis B virus infection and meets the clinical diagnostic criteria for chronic hepatitis B. The nucleoside or nucleotide analogue therapy refers to treatment using entecavir, tenofovir, tenofovir alafenamide, or other antiviral drugs that can inhibit hepatitis B virus replication. The sustained virological response refers to the patient receiving the above antiviral treatment, and the hepatitis B virus DNA level remaining below the detection limit or meeting clinically set virological control standards, and this state being maintained for a preset time requirement, such as for more than six consecutive months. The data acquisition unit 101 can acquire patient data through manual entry, electronic medical record interface, laboratory information system interface, health management database interface, or file import.

[0014] The data acquired by the acquisition unit 101 includes at least age, sex, quantitative value of hepatitis B surface antigen (HBsAg), albumin level, aspartate aminotransferase (AST) level, upper limit of normal for AST, and platelet count. Age can be the patient's actual age in years on the risk prediction date or the specified examination date; sex can be recorded as male or female, or converted to a preset code within the system. The quantitative value of HBsAg refers to the concentration result obtained through quantitative detection, usually expressed in IU / mL, reflecting the level of hepatitis B virus-related antigens in the patient's body. Albumin level refers to the serum albumin test result, usually expressed in g / L, reflecting liver synthetic function and nutritional status. The AST level refers to the AST test result, usually expressed in U / L; the upper limit of normal for AST refers to the upper limit of the normal reference range corresponding to the laboratory or test method performing the AST test. For example, if a laboratory sets the upper limit of normal for AST to be 40 U / L, this value will be used as the standardized benchmark for subsequent APRI calculations. Platelet count refers to the result of peripheral blood platelet detection, and is usually expressed as a 10-1. 9 / L is used as the unit.

[0015] To ensure the comparability and feasibility of subsequent risk prediction results, the data acquisition unit 101 can perform integrity checks and time consistency checks on the acquired data when forming the target patient dataset. The target patient dataset refers to a collection of data generated by the same patient within the same follow-up node or preset time window after meeting the sustained virological response criteria, which can be directly accessed by the subsequent parameter generation unit 102. The preset time window can be set according to the clinical follow-up frequency, for example, using the risk prediction date as a benchmark and selecting the most recent valid laboratory test result; when the test dates for different indicators are not exactly the same, test results within a 30-day or 90-day interval can be selected as the same target patient dataset. If the quantitative value of hepatitis B surface antigen, albumin value, AST value, or platelet count is missing, the data acquisition unit 101 can prompt for supplementation, retesting, or mark the patient as temporarily unsuitable for this risk prediction.

[0016] For example, for a male patient receiving entecavir treatment and whose HBV DNA was below the detection limit for one consecutive year, the predicted age obtained by collection unit 101 was 55 years old, with a hepatitis B surface antigen quantitative value of 850 IU / mL, an albumin value of 42 g / L, an AST value of 36 U / L, an upper limit of normal for AST of 40 U / L, and a platelet count of 150 × 10⁻⁶. 9If / L, the acquisition unit 101 will compile the above information along with the patient's ongoing virological response status into a target patient dataset for that patient. This target patient dataset serves as the data basis for subsequent calculations of hepatitis B surface antigen risk parameters, APRI values, APRI risk parameters, baseline risk scores, and comprehensive risk scores.

[0017] The parameter generation unit 102 is used to generate hepatitis B surface antigen risk parameters based on the quantitative value of hepatitis B surface antigen in the target patient dataset, and to calculate APRI value and APRI risk parameters based on aspartate aminotransferase value, upper limit of normal aspartate aminotransferase value and platelet count.

[0018] In this embodiment, the parameter generation unit 102 is used to read the quantitative value of hepatitis B surface antigen, the value of aspartate aminotransferase, the upper limit of normal value of aspartate aminotransferase, and the platelet count from the target patient dataset formed by the acquisition unit 101, and convert the above raw test values ​​into structured risk parameters that can be directly called by the subsequent scoring unit 104. Here, risk parameters refer to the standardized results formed by converting raw clinical test values ​​according to preset quantification rules. These standardized results may include raw values, grade values, risk level identifiers, and corresponding parameter codes. For example, risk levels can be divided into low-risk, medium-risk, and high-risk levels, and the parameter codes can be recorded as 0, 1, and 2, respectively. Other coding methods can also be adopted based on actual clinical data verification results, as long as consistency is maintained within the same system.

[0019] For the hepatitis B surface antigen (HBsAg) risk parameter, the parameter generation unit 102 first reads the HBsAg quantitative value from the target patient dataset and confirms whether the unit of the value meets the preset unit requirements. If there are differences in the units output by different detection systems, the parameter generation unit 102 can first unify the units; if the detection result is lower than the detection limit, it can be recorded as the detection limit value or processed according to the preset low value code; if the detection result exceeds the detection limit, it can be recorded as the detection limit value or a suggestion can be made to use the quantitative result after dilution and retesting. After completing the numerical unification, the parameter generation unit 102 generates the HBsAg risk parameter according to the preset HBsAg quantification rules. The preset HBsAg quantification rules can adopt an interval grading method. For example, the case where the HBsAg quantitative value is less than 100 IU / mL is determined as low level and assigned parameter code 0, the case between 100 IU / mL and 1000 IU / mL is determined as medium level and assigned parameter code 1, and the case greater than 1000 IU / mL is determined as high level and assigned parameter code 2. The aforementioned thresholds can be preset based on historical follow-up data of the target population, reference ranges of hospital testing platforms, or clinical research results, and stored in the system's parameter configuration table. In one embodiment, the hepatitis B surface antigen risk parameter not only includes the aforementioned parameter codes, but can also retain the original quantitative value of hepatitis B surface antigen and the corresponding risk level, so that the subsequent imbalance correction unit 103 can compare the level differences.

[0020] For the APRI value, parameter generation unit 102 reads the aspartate aminotransferase (AST) value, the upper limit of normal for AST, and the platelet count from the target patient dataset. APRI refers to the aspartate aminotransferase-to-platelet count ratio, which is used in this system to characterize the risk status associated with liver fibrosis. During calculation, parameter generation unit 102 first divides the AST value by the upper limit of normal for AST to obtain a standardized enzyme ratio, then divides this ratio by the platelet count, and multiplies it by a preset amplification factor to obtain an APRI value that is easy to grade. The platelet count is typically calculated using a 10-1 ratio. 9 Using / L as the unit, the preset magnification factor is typically set to 100 to ensure the calculation results are consistent with the clinically commonly used APRI standard. For example, a patient's aspartate aminotransferase (AST) level is 36 U / L, the upper limit of normal for AST is 40 U / L, and the platelet count is 150 × 10⁻⁶. 9 If the value is / L, then parameter generation unit 102 first obtains an enzymatic ratio of 0.9, then divides 0.9 by 150 and multiplies by 100 to obtain an APRI value of 0.6. This process can eliminate the influence of differences in the upper limit of normal in different laboratories on the interpretation of AST results, making APRI values ​​comparable between different patients.

[0021] After obtaining the APRI value, the parameter generation unit 102 generates APRI risk parameters according to preset APRI quantification rules. The preset APRI quantification rules can adopt a range-based grading method. For example, APRI values ​​less than 0.5 are classified as low-level and assigned parameter code 0; values ​​between 0.5 and 1.5 are classified as medium-level and assigned parameter code 1; and values ​​greater than 1.5 are classified as high-level and assigned parameter code 2. These grading thresholds can be calibrated based on sample data from the system's applicable population, or they can be pre-configured by medical institutions according to their follow-up strategies. Continuing with the above patient as an example, if their APRI value is 0.6, the parameter generation unit 102 classifies it as medium-level and generates APRI risk parameters including the original APRI value of 0.6, the APRI risk level "medium-level," and parameter code 1.

[0022] To ensure the stability of the parameter generation results, the parameter generation unit 102 can perform anomaly verification on the input values. For example, if the normal upper limit of aspartate aminotransferase is empty, zero, or obviously unreasonable, the system can call the default normal upper limit or prompt for supplementation of the laboratory reference range; when the platelet count is empty or less than or equal to zero, the system does not perform APRI calculation and marks the target patient dataset as lacking necessary parameters; when the difference between the quantitative value of hepatitis B surface antigen, AST value, or platelet count and adjacent test records exceeds a preset proportion, the system can retain the calculation results and add an anomaly mark for subsequent manual review. Through the above processing, the parameter generation unit 102 can convert the virological indicators and fibrosis-related indicators in the target patient dataset into hepatitis B surface antigen risk parameters and APRI risk parameters, respectively, and provide them as inputs under the same quantitative scale to the imbalance correction unit 103 and the scoring unit 104, thereby supporting subsequent comparison of risk level differences, generation of basic risk scores, and correction of comprehensive risk scores.

[0023] Furthermore, the parameter generation unit is specifically used for: Read the preset reference threshold package for sustained virological response population, which includes the 33rd percentile threshold, 67th percentile threshold, upper reference boundary of the highest risk interval and lower reference boundary of the lowest risk interval for the quantitative value of hepatitis B surface antigen, as well as the 33rd percentile threshold, 67th percentile threshold, upper reference boundary of the highest risk interval and lower reference boundary of the lowest risk interval for the APRI value; The detection boundary processing is performed on the quantitative values ​​of hepatitis B surface antigen in the target patient dataset to obtain the effective quantitative value of hepatitis B surface antigen. Based on the comparison results between the effective quantitative value of hepatitis B surface antigen and the 33rd percentile threshold and the 67th percentile threshold of the quantitative value of hepatitis B surface antigen in the reference threshold package for the sustained virological response population, the risk level of hepatitis B surface antigen, the level code of hepatitis B surface antigen and the classification interval to which hepatitis B surface antigen belongs are generated. The APRI value is calculated based on the aspartate aminotransferase (AST) value, the upper limit of normal aspartate aminotransferase (AST) value, and the platelet count. Based on the comparison between the APRI value and the 33rd percentile threshold and the 67th percentile threshold of the APRI value in the reference threshold package for sustained virological response population, the APRI risk level, APRI level code, and the APRI classification interval are generated. When the effective hepatitis B surface antigen quantitative value or APRI value is in the corresponding highest risk interval or lowest risk interval, the upper reference boundary of the highest risk interval or the lower reference boundary of the lowest risk interval in the reference threshold package for the sustained virological response population is called to supplement the grade interval to which the hepatitis B surface antigen belongs or the grade interval to which the APRI belongs, so that it has a definite upper and lower boundary. The effective quantitative value of hepatitis B surface antigen, the risk level of hepatitis B surface antigen, the grade code of hepatitis B surface antigen, and the grade interval to which hepatitis B surface antigen belongs are packaged into hepatitis B surface antigen risk parameters, and the APRI value, APRI risk level, APRI grade code, and the grade interval to which APRI belongs are packaged into APRI risk parameters.

[0024] In this embodiment, the parameter generation unit converts the quantitative value of hepatitis B surface antigen and the test data related to APRI into risk parameters that can be directly accessed by the subsequent imbalance correction unit and scoring unit after the target patient dataset is formed. To ensure that the parameter grading is applicable to chronic hepatitis B patients who have achieved sustained virological response after nucleoside or nucleotide analog therapy, the parameter generation unit does not directly use fixed thresholds for the general chronic hepatitis B population, but instead reads a preset reference threshold package for the sustained virological response population. This reference threshold package for the sustained virological response population refers to a set of threshold configurations pre-formed based on historical data, follow-up data, or a validated local database of similar treated patients who have achieved sustained virological response. It includes at least the 33rd percentile threshold, the 67th percentile threshold, the upper reference boundary of the highest risk interval, and the lower reference boundary of the lowest risk interval for the quantitative value of hepatitis B surface antigen, as well as the 33rd percentile threshold, the 67th percentile threshold, the upper reference boundary of the highest risk interval, and the lower reference boundary of the lowest risk interval for the APRI value. The 33rd percentile threshold refers to the percentage of corresponding indicator values ​​in the reference population that are not higher than this threshold; the 67th percentile threshold refers to the percentage of corresponding indicator values ​​in the reference population that are not higher than this threshold. Using this reference threshold package ensures that the grading standards for hepatitis B surface antigen quantitative values ​​and APRI values ​​are consistent with the applicable population of this system.

[0025] The parameter generation unit first performs detection boundary processing on the hepatitis B surface antigen (HBsAg) quantitative values ​​in the target patient dataset. This detection boundary processing refers to the numerical correction or labeling of the lower and upper detection limits in the testing equipment or reagent report. For example, when the HBsAg quantitative value is below the lower detection limit, the lower detection limit can be used as the effective HBsAg quantitative value, or the value can be marked as below the lower detection limit and a preset minimum calculable value can be used. When the HBsAg quantitative value is above the upper detection limit, the quantitative result after dilution and retesting can be used as the effective HBsAg quantitative value. If dilution and retesting results are not immediately available, the upper detection limit can be used with an additional "to be verified" label. The value obtained after the above processing is called the effective HBsAg quantitative value, which is used for subsequent risk level generation.

[0026] The parameter generation unit compares the effective quantitative value of hepatitis B surface antigen (HBsAg) with the 33rd and 67th percentile thresholds of HBsAg quantitative values ​​in the reference threshold package for sustained virological response populations, thereby generating an HBsAg risk level, an HBsAg level code, and the HBsAg grading interval. In one possible implementation, when the effective HBsAg quantitative value is lower than the 33rd percentile threshold, a low-risk HBsAg level is generated, and the HBsAg level code is set to 0; when the effective HBsAg quantitative value is not lower than the 33rd percentile threshold but lower than the 67th percentile threshold, a medium-risk HBsAg level is generated, and the HBsAg level code is set to 1; when the effective HBsAg quantitative value is not lower than the 67th percentile threshold, a high-risk HBsAg level is generated, and the HBsAg level code is set to 2. The HBsAg grading interval refers to the numerical range into which the effective HBsAg quantitative value falls, such as a low-risk interval, a medium-risk interval, or a high-risk interval. To avoid boundary values ​​falling into two intervals simultaneously, the system can pre-adopt an interval rule of "lower boundary includes, upper boundary does not include"; for the highest risk interval, since its upper boundary may be infinite in natural expression, the system supplements the upper boundary of the interval with the upper reference boundary of the highest risk interval; for the lowest risk interval, the system supplements the lower boundary of the interval with the lower reference boundary of the lowest risk interval.

[0027] The parameter generation unit also calculates the APRI value based on the aspartate aminotransferase (AST) value, the upper limit of normal (UPR) for AST, and the platelet count in the target patient dataset. Specifically, the parameter generation unit first standardizes the AST value using the UPR to correct for differences caused by varying reference ranges from different laboratories, and then combines this with the platelet count to obtain the APRI value. Platelet counts are typically calculated using a 10-1... 9 / L is used as the unit; if the platelet count is empty or not greater than zero, the parameter generation unit will not generate an APRI value and may output a data shortage warning. For example, a patient's aspartate aminotransferase (AST) value is 36 U / L, the upper limit of normal for AST is 40 U / L, and the platelet count is 150 × 10⁻⁶. 9 If the value is / L, then the aspartate aminotransferase value is equivalent to 0.9 times the upper limit of normal. After combining the platelet count, the APRI value can be calculated to be approximately 0.6.

[0028] After obtaining the APRI value, the parameter generation unit compares the APRI value with the 33rd and 67th percentile thresholds of the APRI value in the reference threshold package for sustained virological response populations, generating the APRI risk level, APRI level code, and the APRI classification interval. Its classification logic is consistent with the hepatitis B surface antigen risk level: a value below the 33rd percentile threshold is generated as a low-risk APRI level and encoded as 0; a value not lower than the 33rd percentile threshold but lower than the 67th percentile threshold is generated as a medium-risk APRI level and encoded as 1; and a value not lower than the 67th percentile threshold is generated as a high-risk APRI level and encoded as 2. The APRI classification interval is also determined by the 33rd and 67th percentile thresholds of the APRI value, as well as the upper reference boundary of the highest-risk interval and the lower reference boundary of the lowest-risk interval, giving the APRI classification interval definite upper and lower boundaries.

[0029] For example, in the pre-defined reference threshold package for sustained virological response populations, the 33rd percentile threshold for HBsAg quantification is 120 IU / mL, the 67th percentile threshold is 950 IU / mL, the lower reference boundary of the lowest risk interval is 0 IU / mL, and the upper reference boundary of the highest risk interval is 10000 IU / mL; the 33rd percentile threshold for APRI value is 0.45, the 67th percentile threshold is 1.20, the lower reference boundary of the lowest risk interval is 0, and the upper reference boundary of the highest risk interval is 3.00. If a patient's effective HBsAg quantification value is 850 IU / mL, it falls between 120 IU / mL and 950 IU / mL, corresponding to a HBsAg risk level, with the HBsAg level code being 1, and the HBsAg grading interval being 120 IU / mL to 950 IU / mL. If a patient's APRI value is 0.60, it falls within the range of 0.45 to 1.20, corresponding to a risk level of 0.45. The APRI level code is 1, and the APRI grading interval is 0.45 to 1.20. For example, if another patient's effective hepatitis B surface antigen (HBsAg) quantitative value is 1500 IU / mL, it is not lower than the 67th percentile threshold, belonging to the high-risk level for HBsAg. The system uses 950 IU / mL as the lower boundary of this high-risk interval and 10000 IU / mL as the upper boundary, thus forming a grading interval for HBsAg with defined boundaries, used for subsequent interval position value calculations.

[0030] Finally, the parameter generation unit encapsulates the effective quantitative value of hepatitis B surface antigen (HBsAg), HBsAg risk level, HBsAg level code, and the HBsAg grading interval into HBsAg risk parameters, and encapsulates the APRI value, APRI risk level, APRI level code, and APRI grading interval into APRI risk parameters. This encapsulation refers to outputting the original or calculated value, risk level, level code, and interval boundary corresponding to the same indicator as a set of structured data. Through this method, the parameter generation unit can generate HBsAg risk parameters and APRI risk parameters applicable to this system according to the reference standard for sustained virological response populations, and provides a clear data basis for the subsequent imbalance correction unit to determine risk level differences, calculate interval position values, and generate virus-fibrosis imbalance correction coefficients.

[0031] Imbalance correction unit 103 is used to compare the risk level difference between the hepatitis B surface antigen risk parameter and the APRI risk parameter, and to generate a virus-fibrosis imbalance correction coefficient when the risk level corresponding to the hepatitis B surface antigen risk parameter is inconsistent with the risk level corresponding to the APRI risk parameter.

[0032] In this embodiment, the imbalance correction unit 103 receives the hepatitis B surface antigen (HBsAg) risk parameter and the APRI risk parameter output by the parameter generation unit 102, and compares whether the corresponding risk levels are consistent. The HBsAg risk parameter is mainly used to characterize the virologically related residual risk that still exists after nucleoside or nucleotide analog treatment and achieving a sustained virological response. The APRI risk parameter is mainly used to characterize the liver fibrosis-related risk reflected by aspartate aminotransferase (AST) and platelet count. For treated chronic hepatitis B patients, the virologically related residual risk and the liver fibrosis-related risk do not always change synchronously. For example, some patients may still have a high HBsAg quantitative value but a low APRI value; conversely, there may be cases where the HBsAg quantitative value is low but the APRI value is high. The imbalance correction unit 103 identifies the above-mentioned inconsistent risk levels and converts the inconsistent state into a virus-fibrosis imbalance correction coefficient, which is then used by the scoring unit 104 to correct the baseline risk score.

[0033] In practical implementation, the imbalance correction unit 103 first reads the hepatitis B surface antigen risk level from the hepatitis B surface antigen risk parameter and the APRI risk level from the APRI risk parameter. The risk level can be low, medium, or high, or other grading systems based on preset rules of the medical institution. For ease of calculation, in one embodiment, the system converts low, medium, and high levels into level codes zero, one, and two, respectively. The imbalance correction unit 103 determines the magnitude and direction of the difference between the hepatitis B surface antigen risk level code and the APRI risk level code by comparing them. The magnitude of the difference refers to the number of levels between the two level codes; the direction of the difference refers to whether the hepatitis B surface antigen risk level is higher than the APRI risk level, or vice versa. If the two level codes are the same, it indicates that their risk levels are consistent and there is no virus-fibrosis imbalance; if the two level codes are different, it indicates that a virus-fibrosis imbalance exists.

[0034] The virus-fibrosis imbalance correction coefficient is a correction parameter used to reflect the degree of inconsistency between the risk levels of the hepatitis B surface antigen (HBsAg) risk parameter and the APRI risk parameter. This correction coefficient can be generated using a preset correction table. To ensure logical consistency in the correction, the preset correction table should meet the monotonicity requirement under the same direction of difference; that is, under the same direction of difference, the greater the difference in risk levels, the greater the correction coefficient should be, and the smaller the difference should be. When the risk levels of the two are the same, the correction coefficient is one, indicating that no additional correction is performed. For example, in one possible implementation, when the HBsAg risk level is one level higher than the APRI risk level, the virus-fibrosis imbalance correction coefficient is 1.5%; when the HBsAg risk level is two levels higher than the APRI risk level, the virus-fibrosis imbalance correction coefficient is 1.10%; when the APRI risk level is one level higher than the HBsAg risk level, the virus-fibrosis imbalance correction coefficient is 1.8%; and when the APRI risk level is two levels higher than the HBsAg risk level, the virus-fibrosis imbalance correction coefficient is 1.15%. Therefore, on the side with a higher risk level for hepatitis B surface antigen, the correction coefficient increases from 1.5% to 1.10% as the risk level increases from level one to level two; on the side with a higher risk level for APRI, the correction coefficient increases from 1.8% to 1.15% as the risk level increases from level one to level two, thus ensuring that the greater the difference in the same direction, the greater the correction magnitude or at least not the decrease.

[0035] The correction magnitudes for the two directions can differ because the hepatitis B surface antigen (HBsAg) risk parameter and the APRI risk parameter represent different sources of risk. When the HBsAg risk level is higher than the APRI risk level, the system considers the patient to have a relatively prominent risk of virologically related residual disease; when the APRI risk level is higher than the HBsAg risk level, the system considers the patient to have a relatively prominent risk of liver fibrosis. Both states require correction of the baseline risk score, but the specific correction magnitude can be determined based on historical follow-up data of treated chronic hepatitis B patients, clinical management strategies, or pre-defined expert rules. Regardless of the specific values ​​used, the pre-defined correction table should remain monotonic in the same direction of difference to avoid situations where the risk level difference is large but the correction magnitude is small.

[0036] For example, if a patient's hepatitis B surface antigen (HBsAg) risk parameter corresponds to a high level and their APRI risk parameter corresponds to a medium level, then the difference between the two is one level, and the direction of the difference is that the HBsAg risk level is higher than the APRI risk level. The imbalance correction unit 103 can generate a virus-fibrosis imbalance correction coefficient of 1.5%. If another patient's HBsAg risk parameter corresponds to a high level and their APRI risk parameter corresponds to a low level, then the difference between the two is two levels, and the HBsAg risk level is still higher than the APRI risk level. The imbalance correction unit 103 can generate a virus-fibrosis imbalance correction coefficient of 1.10%. Since the correction coefficient corresponding to a two-level difference is greater than the correction coefficient corresponding to a one-level difference, this correction process can reflect the increased degree of imbalance between virologically related residual risk and fibrosis-related risk.

[0037] For example, if a patient's hepatitis B surface antigen (HBsAg) risk parameter corresponds to a low level and their APRI risk parameter corresponds to a medium level, then the difference between the two is one level, and the direction of the difference is that the APRI risk level is higher than the HBsAg risk level. The imbalance correction unit 103 can generate a virus-fibrosis imbalance correction coefficient of 1.8%. If the HBsAg risk parameter corresponds to a low level and the APRI risk parameter corresponds to a high level, then the difference between the two is two levels, and the imbalance correction unit 103 can generate a virus-fibrosis imbalance correction coefficient of 1.5%. If both the HBsAg risk parameter and the APRI risk parameter correspond to a medium level, then there is no difference in risk level between the two, and the imbalance correction unit 103 generates a virus-fibrosis imbalance correction coefficient of 1. The scoring unit 104 can then avoid additional upward adjustment of the baseline risk score due to the imbalance.

[0038] In another implementation, the imbalance correction unit 103 may not directly use a fixed numerical table, but instead generate virus-fibrosis imbalance correction coefficients using configuration rules that satisfy monotonic constraints. Specifically, the system can preset a first-level and a second-level difference correction range for the direction where the hepatitis B surface antigen risk level is higher than the APRI risk level, requiring that the second-level difference correction range is not less than the first-level difference correction range; simultaneously, it can preset a first-level and a second-level difference correction range for the direction where the APRI risk level is higher than the hepatitis B surface antigen risk level, requiring that the second-level difference correction range is not less than the first-level difference correction range. Through this configuration method, even if the medical institution adjusts the specific correction values ​​based on local patient data, the basic logical consistency of the imbalance correction unit 103 will not be disrupted.

[0039] To ensure the stability and reliability of the correction results, the imbalance correction unit 103 can also process the data anomaly markers output by the parameter generation unit 102. When there are missing, significantly abnormal, or pending verification markers for the quantitative value of hepatitis B surface antigen, the value of aspartate aminotransferase, the upper limit of normal aspartate aminotransferase, or the platelet count, the imbalance correction unit 103 may temporarily not generate a correction coefficient greater than one, but instead generate a pending verification status or use one as a conservative correction coefficient. If the system has data from adjacent follow-up nodes, the imbalance correction unit 103 can also use the complete correction coefficient when the same imbalance direction occurs consecutively, and use a lower correction coefficient when it occurs only once and the magnitude is small. In this way, the imbalance correction unit 103 can, while maintaining the self-consistency and repeatability of the correction table, clearly convert the grade difference between the hepatitis B surface antigen risk parameter and the APRI risk parameter into a virus-fibrosis imbalance correction coefficient, and output this correction coefficient to the scoring unit 104 for correcting the baseline risk score.

[0040] Furthermore, the imbalance correction unit is specifically used for: Read the quantitative value of hepatitis B surface antigen, the risk level of hepatitis B surface antigen and its corresponding classification interval from the hepatitis B surface antigen risk parameters, and read the APRI value, APRI risk level and its corresponding classification interval from the APRI risk parameters; The risk levels of hepatitis B surface antigen and APRI are converted into level codes, and an initial directed imbalance value is generated based on the difference between the two level codes. The viral side interval position value is generated based on the position of the quantitative value of hepatitis B surface antigen in its respective grade interval, and the fibrosis side interval position value is generated based on the position of the APRI value in its respective grade interval. When the initial directed imbalance value indicates that the risk level of hepatitis B surface antigen is higher than the risk level of APRI, the virus-side interval position value and the fibrosis-side interval position value are used to determine whether the initial directed imbalance value belongs to the virus-dominant type of credible imbalance; when the initial directed imbalance value indicates that the risk level of APRI is higher than the risk level of hepatitis B surface antigen, the fibrosis-side interval position value and the virus-side interval position value are used to determine whether the initial directed imbalance value belongs to the fibrosis-dominant type of credible imbalance. When the credible imbalance is either virologically dominant or fibroticly dominant, a virological-fibrotic imbalance correction coefficient is generated based on the direction and absolute value of the initial directed imbalance value. When the credible imbalance is not virologically dominant, the absolute value of the initial directed imbalance value is reduced by one level to generate the virological-fibrotic imbalance correction coefficient.

[0041] In this embodiment, the imbalance correction unit is used to further determine whether the difference in risk levels between the hepatitis B surface antigen (HBsAg) risk parameter and the APRI risk parameter constitutes a stable and reliable true imbalance, rather than an accidental difference caused by test values ​​approaching the grading boundary. The HBsAg risk parameter may include the HBsAg quantitative value, the HBsAg risk level, and its corresponding grading interval. The APRI risk parameter may include the APRI value, the APRI risk level, and its corresponding grading interval. The corresponding grading interval refers to the numerical range used to generate the corresponding risk level. For example, an HBsAg quantitative value less than 100 IU / mL corresponds to a low-risk interval, 100 IU / mL to 1000 IU / mL corresponds to a medium-risk interval, and greater than 1000 IU / mL corresponds to a high-risk interval; an APRI value less than 0.5 corresponds to a low-risk interval, 0.5 to 1.5 corresponds to a medium-risk interval, and greater than 1.5 corresponds to a high-risk interval. For the open interval corresponding to the highest risk level, the system can pre-set the upper reference boundary. For example, the upper reference boundary for the high-risk interval of hepatitis B surface antigen can be set to 10,000 IU / mL, and the upper reference boundary for the high-risk interval of APRI can be set to 3.0, so as to calculate its position within the interval. For the open interval corresponding to the lowest risk level, the detection limit or zero value can be used as the lower reference boundary.

[0042] The imbalance correction unit first reads the hepatitis B surface antigen (HBsAg) risk level and the APRI risk level, and converts them into level codes under the same level system. In one implementation, low-risk, medium-risk, and high-risk levels correspond to 0, 1, and 2, respectively. Subsequently, the imbalance correction unit subtracts the APRI risk level code from the HBsAg risk level code to obtain an initial directed imbalance value. The sign of this initial directed imbalance value indicates the direction of the imbalance: a positive value indicates that the HBsAg risk level is higher than the APRI risk level, and a negative value indicates that the APRI risk level is higher than the HBsAg risk level. The absolute value represents the number of levels between the two. For example, when the HBsAg risk level is high and the APRI risk level is medium, the initial directed imbalance value is positive level one; when the HBsAg risk level is low and the APRI risk level is high, the initial directed imbalance value is negative level two.

[0043] To avoid mechanical correction based solely on grading codes, the imbalance correction unit further calculates the viral-side interval position of the hepatitis B surface antigen (HBsAg) quantitative value within its respective grading interval, and the fibrotic-side interval position of the APRI value within its respective grading interval. The interval position value refers to the proportion of the target value relative to the lower and upper boundaries within its grading interval, and can be set to a range of 0 to 1. A value closer to 0 indicates proximity to the lower boundary of the interval, and a value closer to 1 indicates proximity to the upper boundary. For example, if a patient's HBsAg quantitative value is 800 IU / mL, and its intermediate-risk interval is 100 IU / mL to 1000 IU / mL, then its distance from the lower boundary is 700 IU / mL, the interval width is 900 IU / mL, and the viral-side interval position value is approximately 0.78, indicating that the HBsAg quantitative value is close to the upper boundary within the intermediate-risk interval. For example, if a patient's APRI value is 0.55, which falls within the intermediate risk range of 0.5 to 1.5, then the value of the fibrotic side interval is approximately 0.05, indicating that although the APRI value is classified as intermediate risk, it is very close to the lower boundary of the intermediate risk range.

[0044] When the initial directed imbalance value indicates that the hepatitis B surface antigen (HBsAg) risk level is higher than the APRI risk level, the imbalance correction unit considers this situation as a candidate for a virus-dominant imbalance and uses the virus-side interval position value and the fibrosis-side interval position value to determine whether it belongs to a credible virus-dominant imbalance. In one implementation, a credible virus-dominant imbalance is determined when the virus-side interval position value is not less than 0.60 and the fibrosis-side interval position value is not greater than 0.40. This means that the HBsAg quantitative value on the higher-risk side has not just crossed the grading boundary, but is relatively stable and close to the high-value side within its grading interval; at the same time, the APRI value on the lower-risk side is not close to the boundary of rising to a higher level. Conversely, if the HBsAg quantitative value has just entered a higher-risk level, or the APRI value is already close to the upper boundary of the next level, the imbalance may be caused by boundary fluctuations, and the imbalance correction unit does not treat it as a completely credible imbalance.

[0045] When the initial directed imbalance value indicates that the APRI risk level is higher than the hepatitis B surface antigen risk level, the imbalance correction unit considers this situation as a candidate for a fibrosis-dominant imbalance and uses the fibrosis-side interval position value and the virus-side interval position value to determine whether it belongs to a fibrosis-dominant credible imbalance. In one embodiment, when the fibrosis-side interval position value is not less than 0.60 and the virus-side interval position value is not greater than 0.40, it is determined to be a fibrosis-dominant credible imbalance. This means that the APRI value on the higher-risk side has a relatively stable high value characteristic within its interval, while the hepatitis B surface antigen quantitative value is not in a critical state of about to rise to a higher risk level.

[0046] When determining whether a credible imbalance belongs to either a virus-dominant or fibrosis-dominant type, the imbalance correction unit generates a virus-fibrosis imbalance correction coefficient according to the direction and absolute value of the initial directional imbalance value. For example, if the hepatitis B surface antigen risk level is higher than the APRI risk level by one level and constitutes a credible imbalance, a correction coefficient of 1.05 can be generated; if it is two levels higher, a correction coefficient of 1.10 can be generated. If the APRI risk level is higher than the hepatitis B surface antigen risk level by one level and constitutes a credible imbalance, a correction coefficient of 1.08 can be generated; if it is two levels higher, a correction coefficient of 1.15 can be generated. The correction coefficients in the same direction do not decrease as the level difference increases, thus ensuring the consistency of the correction logic.

[0047] When the imbalance is not credible, the imbalance correction unit reduces the absolute value of the initial directed imbalance by one level before generating the virus-fibrosis imbalance correction coefficient. For example, if the initial directed imbalance is positive level II, but the hepatitis B surface antigen (HBsAg) quantitative value is only slightly higher than the lower boundary of the high-risk interval, or the APRI value is close to the upper boundary of its lower-level interval, the system reduces the imbalance order from level II to level I, and then generates a correction coefficient according to the case of a higher HBsAg risk level. If the initial directed imbalance is only level I and is not credible, the imbalance is reduced by one level to level zero, and the imbalance correction unit generates 1 as the correction coefficient, indicating that no additional correction is performed. In this way, the imbalance correction unit not only compares the level difference between the HBsAg risk parameter and the APRI risk parameter, but also combines the position of their original values ​​in their respective grading intervals to determine whether the level difference is stable and reliable, thereby avoiding the erroneous amplification of random fluctuations near the threshold value, and converting the credible imbalance relationship between virological residual risk and fibrosis-related risk into a virus-fibrosis imbalance correction coefficient that can be used by the scoring unit.

[0048] Scoring unit 104 is used to generate a baseline risk score based on age, gender, albumin level, hepatitis B surface antigen risk parameter, and APRI risk parameter, and then corrects the baseline risk score based on the virus-fibrosis imbalance correction coefficient to obtain a comprehensive risk score.

[0049] In this embodiment, the scoring unit 104 receives the target patient dataset, hepatitis B surface antigen risk parameters, APRI risk parameters, and the virus-fibrosis imbalance correction coefficient output by the imbalance correction unit 103, and converts the above data into a comprehensive risk score that can be further processed by the probability conversion unit 105. The baseline risk score refers to the initial score obtained after assigning scores based on the patient's age, gender, albumin level, hepatitis B surface antigen risk parameters, and APRI risk parameters, before considering the imbalance relationship between the hepatitis B surface antigen risk parameters and APRI risk parameters. The comprehensive risk score refers to the score obtained after correcting the baseline risk score using the virus-fibrosis imbalance correction coefficient, which characterizes the patient's overall risk level of developing hepatocellular carcinoma at the current follow-up point.

[0050] Scoring unit 104 can pre-store a scoring rule table, which includes age, gender, albumin, hepatitis B surface antigen (HBsAg) and APRI (Audio-Primary Intake) scoring items. The age scoring item can be assigned points according to age ranges, for example, zero points for those under 40, one point for those between 40 and 54, two points for those between 55 and 64, and three points for those over 65. The gender scoring item can be set to zero points for females and one point for males. The albumin scoring item can be assigned points according to liver function status, for example, zero points for albumin levels not lower than 40 g / L, one point for levels between 35 g / L and lower than 40 g / L, and two points for levels below 35 g / L. The HBsAg scoring item can be assigned zero, one, or two points based on the low, medium, and high risk levels corresponding to the HBsAg risk parameter, respectively. The APRI scoring item can be assigned zero, one, or two points based on the low, medium, and high risk levels corresponding to the APRI risk parameter, respectively. The above scores and ranges are only one possible implementation method. Medical institutions may also adjust them based on historical follow-up data of patients with chronic hepatitis B who have been treated. However, they should be fixed in advance within the same system to ensure that the same patient data can obtain consistent scoring results.

[0051] When generating the baseline risk score, scoring unit 104 first reads the age, gender, and albumin value from the target patient dataset, and then reads the hepatitis B surface antigen risk parameter and APRI risk parameter from parameter generation unit 102. It then looks up the corresponding individual scores in the scoring rule table and adds them together to obtain the baseline risk score. For example, if a patient is a 55-year-old male with an albumin value of 42 g / L, a moderate hepatitis B surface antigen risk parameter, and a moderate APRI risk parameter, then age corresponds to 2 points, gender to 1 point, albumin to 0 points, hepatitis B surface antigen risk parameter to 1 point, and APRI risk parameter to 1 point. Scoring unit 104 adds these individual scores together to obtain a baseline risk score of 5 points.

[0052] After obtaining the baseline risk score, scoring unit 104 corrects the baseline risk score according to the virus-fibrosis imbalance correction coefficient. If the virus-fibrosis imbalance correction coefficient is one, it indicates that there is no inconsistency between the hepatitis B surface antigen risk parameter and the APRI risk parameter, and scoring unit 104 can directly use the baseline risk score as the comprehensive risk score. If the virus-fibrosis imbalance correction coefficient is greater than one, it indicates that there is an inconsistency between the virological residual risk and the fibrosis-related risk, and scoring unit 104 can adjust the baseline risk score upward according to the correction coefficient to obtain the comprehensive risk score. For example, if a patient's baseline risk score is six points and the virus-fibrosis imbalance correction coefficient is one and one-tenth, then scoring unit 104 will correct the six points to obtain 6.6 points. The system can retain one decimal place as the comprehensive risk score, or it can process it according to preset rounding rules, such as rounding to seven points or rounding up to seven points, but the rounding method should be pre-configured and kept consistent.

[0053] In another implementation, the scoring unit 104 can also employ different correction methods for different imbalance directions. When the hepatitis B surface antigen (HBsAg) risk parameter is higher than the APRI risk parameter, the scoring unit 104 can increase the weight of the HBsAg scoring item to highlight the residual risk of high HBsAg levels even after sustained virological response; when the APRI risk parameter is higher than the HBsAg risk parameter, the scoring unit 104 can increase the weight of the APRI scoring item to highlight the risk associated with liver fibrosis. Regardless of whether the overall base risk score is corrected or a specific scoring item is weighted, the scoring unit 104 should output a clear comprehensive risk score and transmit this comprehensive risk score to the probability conversion unit 105. Through the above processing, the scoring unit 104 can further incorporate the virus-fibrosis imbalance correction coefficient on top of conventional demographic factors, liver function factors, virological factors, and fibrosis factors, making the final score more consistent with the risk characteristics of treated chronic hepatitis B patients.

[0054] Furthermore, the scoring unit is specifically used for: Read the preset scoring table, and generate age score, gender score, albumin score, hepatitis B surface antigen score and APRI score respectively based on age, gender, albumin value, hepatitis B surface antigen risk parameter and APRI risk parameter, and accumulate them to generate a basic risk score; The imbalance correction pathway is determined based on the relationship between the hepatitis B surface antigen (HBsAg) grade code and the APRI grade code. When the HBsAg grade code is greater than the APRI grade code, it is determined to be the viral residual correction pathway. When the APRI grade code is greater than the HBsAg grade code, it is determined to be the fibrosis pressure correction pathway. When the two are the same, the imbalance correction pathway is closed. Liver reserve gating coefficients were generated based on albumin scores. When albumin scores were 0, 1, and 2, the liver reserve gating coefficients for the viral residual correction channel were 1.00, 0.90, and 0.80, respectively, and the liver reserve gating coefficients for the fibrosis pressure correction channel were 1.00, 1.10, and 1.20, respectively. When the viral residual correction channel is activated, the difference between the hepatitis B surface antigen score and the virus-fibrosis imbalance correction coefficient minus 1 is multiplied and then multiplied by the liver reserve gating coefficient to generate the viral residual increment score, while the fibrosis pressure increment score is set to 0. When the fibrosis pressure correction channel is open, the APRI score is multiplied by the difference between the virus-fibrosis imbalance correction coefficient and the value after subtracting 1, and then multiplied by the liver reserve gating coefficient to generate the fibrosis pressure increment score, while the virus residual increment score is set to 0; when the imbalance correction channel is closed, both the virus residual increment score and the fibrosis pressure increment score are set to 0. The comprehensive risk score is obtained by adding the basic risk score, the viral residual increment score, and the fibrotic stress increment score.

[0055] In this embodiment, the scoring unit integrates the patient's baseline demographic factors, liver reserve status, virological residual risk, fibrosis-related risk, and virus-fibrosis imbalance into a comprehensive risk score after the parameter generation unit and imbalance correction unit output the corresponding results. The preset scoring table refers to the scoring mapping rules pre-stored in the system, used to convert age, gender, albumin level, hepatitis B surface antigen risk parameter, and APRI risk parameter into corresponding scores. This preset scoring table can be determined based on historical follow-up data of chronic hepatitis B patients who have achieved sustained virological response after nucleoside or nucleotide analog treatment, or it can be pre-configured based on validated clinical rules and kept fixed within the same version of the system. To ensure that the direction of imbalance is consistent with the direction of incremental scoring, the hepatitis B surface antigen score in the preset scoring table remains monotonically constant with the hepatitis B surface antigen level code, and the APRI score remains monotonically constant with the APRI level code; that is, the higher the risk level code, the lower the corresponding score.

[0056] The scoring unit first reads a preset scoring table and generates an age score based on the age in the target patient dataset, a gender score based on gender, an albumin score based on albumin levels, a hepatitis B surface antigen (HBsAg) score based on HBsAg risk parameters, and an APRI score based on APRI risk parameters. For example, age can be assigned zero, one, two, and three points respectively for those under 40, 40-54, 55-64, and over 65; gender can be set to zero for females and one for males; albumin levels can be assigned zero, one, and two points respectively for those not less than 40 g / L, 35 g / L to less than 40 g / L, and less than 35 g / L; and HBsAg and APRI risk parameters can be assigned zero, one, and two points respectively for low, medium, and high risk levels. The above scoring rules are one possible implementation; in actual application, adjustments can be made based on the validation queue, but the adjusted HBsAg and APRI scores should still satisfy the constraint that they do not decrease monotonically with the corresponding level code. The scoring unit sums the age score, gender score, albumin score, hepatitis B surface antigen score, and APRI score to generate a baseline risk score. This baseline risk score refers to the initial risk score before considering the direction of imbalance between the hepatitis B surface antigen risk level and the APRI risk level.

[0057] Subsequently, the scoring unit determines the imbalance correction pathway based on the relationship between the hepatitis B surface antigen (HBsAg) risk level code and the APRI (Audio-Primary Indication Level) risk level code. This imbalance correction pathway is a data processing path used to determine the target and direction of correction when an inconsistency is found between the HBsAg risk level and the APRI risk level. When the HBsAg risk level code is greater than the APRI risk level code, it indicates that the risk of virological residue is higher than the risk of fibrosis, and the scoring unit determines that the virological residue correction pathway is activated. When the APRI risk level code is greater than the HBsAg risk level code, it indicates that the risk of fibrosis is higher than the risk of virological residue, and the scoring unit determines that the fibrosis pressure correction pathway is activated. When the two risk level codes are the same, it indicates that the two risk levels are equal, and the scoring unit deactivates the imbalance correction pathway.

[0058] The scoring unit also generates a liver reserve gating coefficient based on the albumin score. This liver reserve gating coefficient is a coefficient used to adjust the magnitude of imbalance correction based on liver reserve status. In the baseline risk score, the albumin score reflects the patient's baseline risk of liver synthetic function, while in the liver reserve gating, it determines which type of imbalance—virological residual risk or fibrosis-related risk—should be strengthened. Specifically, the decrease in liver reserve reflected by decreased albumin is usually strongly correlated with fibrosis progression, cirrhosis tendency, or impaired liver parenchymal function. Therefore, when the APRI risk level is higher than the hepatitis B surface antigen risk level, a higher albumin score results in a larger liver reserve gating coefficient in the fibrosis pressure correction pathway, thus strengthening the role of fibrosis pressure risk in the overall risk score. Conversely, the virological residual risk reflected by the quantitative hepatitis B surface antigen value is not on the same risk axis as the decrease in liver reserve. When liver reserve has already decreased, the risk-dominant direction leans more towards fibrosis and insufficient liver functional reserve. Therefore, in the virological residual correction pathway, a higher albumin score results in a smaller liver reserve gating coefficient, reducing the relative increase of the virological residual axis on the overall risk score. Therefore, the albumin score uses opposite gating in the two channels, which is not a contradictory repetition of the same indicator, but rather a processing logic that expresses the shift of the risk-dominant axis from the viral residual axis to the fibrotic pressure axis when liver reserve declines.

[0059] In one possible implementation, when the albumin score is 0, 1, or 2, if the residual viral load correction channel is enabled, the liver reserve gating coefficients are 1.00, 0.90, and 0.80, respectively; if the fibrosis pressure correction channel is enabled, the liver reserve gating coefficients are 1.00, 1.10, and 1.20, respectively. These values ​​can be stored as a system preset configuration and used consistently within the same version of the system. With this setting, patients with good albumin status retain the full residual viral load increment in the residual viral load correction channel, while patients with poor albumin status have an increased fibrosis pressure increment in the fibrosis pressure correction channel. This makes the overall risk score more consistent with the relative relationship between residual viral load risk, fibrosis risk, and liver reserve status in treated patients with sustained virological responses.

[0060] When the residual virus correction channel is activated, the scoring unit generates a residual virus increment score based on the hepatitis B surface antigen (HBsAg) score, the virus-fibrosis imbalance correction coefficient, and the liver reserve gating coefficient. Specifically, the scoring unit first obtains the increase in the virus-fibrosis imbalance correction coefficient relative to one; for example, if the correction coefficient is 1.10, the increase is 0.10. Then, the HBsAg score is multiplied by this increase, and then multiplied by the liver reserve gating coefficient corresponding to the residual virus correction channel to obtain the residual virus increment score. At this point, since the imbalance direction is determined to be viral residual dominance, the fibrosis pressure increment score is set to zero. For example, if a patient has a HBsAg score of 2, an APRI score of 0, a virus-fibrosis imbalance correction coefficient of 1.10 generated by the imbalance correction unit, and an albumin score of 0, the liver reserve gating coefficient of the residual virus correction channel is 1.00. Therefore, the residual virus increment score is 2 multiplied by 0.10 and then multiplied by 1.00, resulting in 0.20, and the fibrosis pressure increment score is zero.

[0061] When the fibrosis pressure correction channel is activated, the scoring unit generates an incremental fibrosis pressure score based on the APRI score, the virus-fibrosis imbalance correction coefficient, and the liver reserve gating coefficient. Specifically, the scoring unit also obtains the increase in the virus-fibrosis imbalance correction coefficient relative to one, then multiplies the APRI score by this increase, and finally multiplies it by the liver reserve gating coefficient corresponding to the fibrosis pressure correction channel to obtain the incremental fibrosis pressure score. The residual virus increment score is set to zero at this time. For example, if a patient has an APRI score of 2, a hepatitis B surface antigen score of 0, a virus-fibrosis imbalance correction coefficient of 1.15, and an albumin score of 2, then the liver reserve gating coefficient for the fibrosis pressure correction channel is 1.20. Therefore, the incremental fibrosis pressure score is 2 multiplied by 0.15 and then multiplied by 1.20, resulting in 0.36, and the residual virus increment score is zero.

[0062] When the imbalance correction channel is closed, it indicates that the hepatitis B surface antigen (HBsAg) grading code is the same as the APRI grading code. The scoring unit sets both the residual viral load (RVV) increment score and the fibrosis pressure increment score to zero, ensuring that the baseline risk score is not additionally corrected due to the imbalance. Finally, the scoring unit adds the baseline risk score, RVV increment score, and fibrosis pressure increment score to obtain a comprehensive risk score, which is then output to the probability conversion unit. The comprehensive risk score can retain at least two decimal places, and the probability conversion unit can directly perform probability conversion or interpolation based on the decimal score, without needing to round it down beforehand in the scoring unit. This ensures that even if the RVV increment score or fibrosis pressure increment score is a decimal, it can still have a calculable impact on the probability of occurrence in subsequent time windows. In this way, the scoring unit does not simply amplify all the basic risk scores as a whole, but selects the corresponding correction channel according to the relative levels of the hepatitis B surface antigen risk parameter and the APRI risk parameter, and combines the albumin score to form a directional liver reserve gating, so that the comprehensive risk score can reflect the relative dominant relationship between the risk of residual virology, the risk of fibrosis pressure and the liver reserve status in treated and continuously virologically responsive chronic hepatitis B patients.

[0063] The probability conversion unit 105 is used to generate multiple time window probabilities based on the comprehensive risk score, including the probability of occurrence of hepatocellular carcinoma in three years, five years, and seven years.

[0064] In this embodiment, the probability conversion unit 105 receives the comprehensive risk score output by the scoring unit 104 and converts the comprehensive risk score into a probability result of the patient developing hepatocellular carcinoma within different follow-up time ranges. The multi-time-window probability refers to providing the likelihood of hepatocellular carcinoma occurrence within three, five, and seven-year time ranges for the same target patient dataset and the same comprehensive risk score. The three-year probability represents the estimated probability of developing hepatocellular carcinoma within three years from the current risk prediction date, the five-year probability represents the estimated probability within five years from the current risk prediction date, and the seven-year probability represents the estimated probability within seven years from the current risk prediction date. The above time ranges can be pre-configured according to clinical follow-up needs, but in this embodiment, three, five, and seven years are preferred to simultaneously reflect short-term, medium-term, and longer-term risks.

[0065] The probability conversion unit 105 can pre-store the probability conversion relationship between the comprehensive risk score and the probability of hepatocellular carcinoma. The probability conversion relationship refers to a preset data processing rule used to map the comprehensive risk score to the probability of occurrence, which can be implemented in the form of a conversion table, segmented conversion rules, or continuous conversion curves. To facilitate implementation by those skilled in the art, in one embodiment, the probability conversion unit 105 uses a conversion table, that is, it pre-stores the three-year, five-year, and seven-year occurrence probabilities corresponding to different comprehensive risk score ranges in the system. For example, when the comprehensive risk score is less than three points, it corresponds to a three-year occurrence probability of less than 1%, a five-year occurrence probability of less than 2%, and a seven-year occurrence probability of less than 3%; when the comprehensive risk score is three to five points, it corresponds to a three-year occurrence probability of 1% to 3%, a five-year occurrence probability of 2% to 5%, and a seven-year occurrence probability of 3% to 8%; when the comprehensive risk score is five to eight points, it corresponds to a three-year occurrence probability of 3% to 8%, a five-year occurrence probability of 5% to 12%, and a seven-year occurrence probability of 8% to 18%; when the comprehensive risk score is greater than eight points, it corresponds to a higher occurrence probability range. The above probability values ​​are merely examples illustrating the probability conversion method. In practical applications, they can be pre-determined based on historical follow-up data of chronic hepatitis B patients who have received treatment and achieved sustained virological response, and stored as a fixed configuration to ensure that the same comprehensive risk score can produce consistent probability outputs.

[0066] In another implementation, if the comprehensive risk score has decimal places, the probability conversion unit 105 can generate the probability of occurrence using interpolation between adjacent score points. This interpolation refers to determining the corresponding probability based on the position of the comprehensive risk score between two stored score points when the comprehensive risk score lies between those two points. For example, the system pre-stores a five-year occurrence probability of 7% for a comprehensive risk score of 6 and 10% for a comprehensive risk score of 7. When a patient's comprehensive risk score is 6.5, the probability conversion unit 105 can determine the five-year occurrence probability as a value between 7% and 10%, such as 8.5%. The three-year and seven-year occurrence probabilities can also be converted using the same method. This method avoids the comprehensive risk score being roughly categorized into a fixed range due to decimal places, improving the continuity and stability of the probability output.

[0067] When performing the conversion, the probability conversion unit 105 can first determine whether the comprehensive risk score is within the preset valid score range. If the comprehensive risk score is lower than the minimum score range, the probability conversion unit 105 can output the probability of occurrence according to the minimum risk range; if the comprehensive risk score is higher than the maximum score range, the probability conversion unit 105 can output the probability of occurrence according to the maximum risk range, and can add a high-risk warning label; if the comprehensive risk score cannot be generated due to missing necessary data, the probability conversion unit 105 will not output the probability of occurrence and will send a data insufficiency label to the risk output unit 106. Through the above-mentioned anomaly handling, invalid scores can be prevented from directly entering the probability conversion process.

[0068] For example, after a patient's comprehensive risk score is processed by scoring unit 104, the probability conversion unit 105, after reading the preset probability conversion relationship, can determine that the patient's probability of developing hepatocellular carcinoma in three years is approximately 5%, in five years it is approximately 9%, and in seven years it is approximately 14%. If another patient's comprehensive risk score is 2.5, the probability conversion unit 105 can output lower probabilities for three, five, and seven years. Thus, the probability conversion unit 105 does not change the comprehensive risk score itself, but rather converts it into a time window probability result that is easier for clinicians to understand and use. This allows the subsequent risk output unit 106 to perform risk stratification and follow-up prompts based on these multiple time window probabilities.

[0069] To improve the traceability of conversion results, the probability conversion unit 105 can also simultaneously record the version of the probability conversion relationship used, the comprehensive risk score, the conversion date, and whether interpolation was used when outputting the probability of occurrence in multiple time windows. This allows for clear identification of the conversion rule used to generate a particular risk prediction result during subsequent follow-ups or system parameter updates, facilitating clinical review and data quality management. Through these settings, the probability conversion unit 105 can stably and definitively convert the comprehensive risk score into the three-year, five-year, and seven-year probabilities of hepatocellular carcinoma occurrence, providing a reliable data foundation for the risk output unit 106 to generate risk stratification results.

[0070] Furthermore, the probability conversion unit is specifically used for: Read the preset conversion table of the incidence rate of the sustained virological response population. The conversion table of the incidence rate of the sustained virological response population includes multiple comprehensive risk score anchor points, and the conditional segmentation incidence rate of 0 to 3 years, conditional segmentation incidence rate of 3 to 5 years and conditional segmentation incidence rate of 5 to 7 years corresponding to each comprehensive risk score anchor point. Among them, the conditional segmentation incidence rate of the same time period does not decrease as the comprehensive risk score anchor point increases, and the cumulative incidence probability obtained by recursively extrapolating the remaining non-occurrence probability does not exceed 1. Based on the comprehensive risk score, adjacent low-score anchor points and adjacent high-score anchor points are determined in the segmented incidence conversion table of the sustained virological response population, and the score position value of the comprehensive risk score between the adjacent low-score anchor points and adjacent high-score anchor points is generated. Based on the score position value, the 0-3 year condition segment incidence rate, 3-5 year condition segment incidence rate, and 5-7 year condition segment incidence rate corresponding to adjacent low score anchor points and adjacent high score anchor points are interpolated in order-preserving manner to generate the 0-3 year condition segment incidence rate, 3-5 year condition segment incidence rate, and 5-7 year condition segment incidence rate of the current patient. Each condition segment incidence rate obtained by interpolation is located between the corresponding condition segment incidence rates of adjacent low score anchor points and adjacent high score anchor points. The incidence rate of conditional segments from 0 to 3 years is used as the probability of hepatocellular carcinoma in three years. The probability of hepatocellular carcinoma in five years is generated by adding the product of the three-year probability of occurrence, the three-year probability of non-occurrence, and the conditional segmentation of the 3- to 5-year incidence rates. The probability of hepatocellular carcinoma in five years is generated by adding the product of the probability of no occurrence in five years and the conditional segmentation of occurrence from 5 to 7 years. The probability of hepatocellular carcinoma in seven years is then generated by adding the probability of hepatocellular carcinoma in three years, five years, and seven years as the multi-time-window occurrence probabilities.

[0071] In this embodiment, the probability conversion unit is used to convert the comprehensive risk score output by the scoring unit into the probability of hepatocellular carcinoma (HCC) occurrence at three, five, and seven years. The segmented incidence conversion table for sustained virological response is a pre-established conversion table based on follow-up data of chronic hepatitis B patients who have achieved sustained virological response after nucleoside or nucleotide analog therapy. It includes multiple comprehensive risk score anchor points and the segmented incidence rate corresponding to each anchor point. The comprehensive risk score anchor points are pre-set scoring nodes, such as three, four, five, and six points, used as reference points for probability conversion. The conditional segmented incidence rate refers to the probability that a patient who has not yet developed HCC at the beginning of a certain time period will develop HCC during that time period. For example, the 3- to 5-year conditional segmented incidence rate is not the total incidence rate over five years from baseline, but rather the probability that a patient who has not developed HCC in the first three years will develop HCC between the third and fifth years. By using conditional segmented incidence rates, the three-year, five-year, and seven-year occurrence probabilities can be recursively formed according to the follow-up time, avoiding treating multiple time window probabilities as isolated results.

[0072] In the conversion table of segmented incidence rates for patients with sustained virological responses, the conditional segmented incidence rate for the same time period does not decrease as the comprehensive risk score anchor point increases. For example, for the conditional segmented incidence rate of 0 to 3 years, the higher the comprehensive risk score anchor point, the lower the corresponding incidence rate is compared to the incidence rate corresponding to the lower score anchor point; the same constraint is applied to the conditional segmented incidence rates of 3 to 5 years and 5 to 7 years. This constraint ensures that patients with high comprehensive risk scores do not have lower incidence rates for the same time period than patients with low comprehensive risk scores. Furthermore, the incidence rates for each segment in the table are pre-defined to meet the requirement that the cumulative probability does not exceed one; that is, after recursively extrapolating based on the remaining non-occurrence probability segment by segment, the cumulative probability of occurrence for three, five, and seven years is between zero and one.

[0073] After receiving the comprehensive risk score, the probability conversion unit first searches for the score interval of the comprehensive risk score in the segmented incidence conversion table for sustained virological response populations, and determines the adjacent low-score anchor point and the adjacent high-score anchor point. The adjacent low-score anchor point refers to the score anchor point that is no higher than the current comprehensive risk score and is closest to it; the adjacent high-score anchor point refers to the score anchor point that is no lower than the current comprehensive risk score and is closest to it. For example, if the conversion table has two adjacent score anchor points, 6 and 7, and a patient's comprehensive risk score is 6.4, then 6 is the adjacent low-score anchor point, and 7 is the adjacent high-score anchor point. Subsequently, the probability conversion unit generates a score position value. The score position value represents the position of the current comprehensive risk score between the adjacent low-score anchor point and the adjacent high-score anchor point, and can take a value from zero to one; in the above example, 6.4 is located within the 6-7 score interval, about 40% close to the low-score anchor point, so the score position value can be 0.4.

[0074] After obtaining the score position value, the probability conversion unit performs order-preserving interpolation on the 0-3 year conditional segmentation incidence rates, 3-5 year conditional segmentation incidence rates, and 5-7 year conditional segmentation incidence rates corresponding to adjacent low-score anchor points and adjacent high-score anchor points, respectively. Order-preserving interpolation means obtaining the current patient's segmentation incidence rate between the corresponding segmentation incidence rates of two adjacent score anchor points based on the score position value, ensuring that the interpolated segmentation incidence rate is not lower than the segmentation incidence rate corresponding to the low-score anchor point and not higher than the segmentation incidence rate corresponding to the high-score anchor point. For example, if the 0-3 year conditional segmentation incidence rate corresponding to a 6-point anchor point is 3%, the 0-3 year conditional segmentation incidence rate corresponding to a 7-point anchor point is 5%, and the current score position value is 0.4, then the current patient's 0-3 year conditional segmentation incidence rate can be determined as 3.8%. If the 3-5 year conditional segmentation incidence rate corresponding to a 6-point anchor point is 4%, and the 3-5 year conditional segmentation incidence rate corresponding to a 7-point anchor point is 6%, then the current patient's 3-5 year conditional segmentation incidence rate can be determined as 4.8%. The 5- to 7-year conditional segmentation incidence rates are also generated in the same way. Through this processing, the decimal part of the comprehensive risk score can be retained and used in probability conversion, without first rounding the comprehensive risk score.

[0075] The probability conversion unit uses the current patient's 0-3 year conditional segmented incidence rate as the three-year probability of hepatocellular carcinoma (HCC) occurrence. Subsequently, the probability conversion unit generates a three-year non-occurrence probability based on the three-year HCC occurrence probability; this three-year non-occurrence probability is the probability of not developing HCC within three years. For example, if the three-year HCC occurrence probability is 3.8%, then the three-year non-occurrence probability is 96.2%. The probability conversion unit then multiplies the three-year non-occurrence probability by the 3-5 year conditional segmented incidence rate to obtain the newly added incidence probability during the third to fifth year, and adds this newly added incidence probability to the three-year HCC occurrence probability to obtain the five-year HCC occurrence probability. For example, if the three-year HCC occurrence probability is 3.8%, and the 3-5 year conditional segmented incidence rate is 4.8%, then the newly added incidence probability during the third to fifth year is 96.2% multiplied by 4.8%, approximately 4.6%, and the five-year HCC occurrence probability is approximately 8.4%.

[0076] After generating the five-year probability of hepatocellular carcinoma (HCC) occurrence, the probability conversion unit further generates the five-year probability of no occurrence. This five-year probability of no occurrence is then superimposed with the conditional segmented incidence rates from years 5 to 7 to obtain the seven-year probability of HCC occurrence. For example, if the five-year probability of HCC occurrence is 8.4%, then the five-year probability of no occurrence is 91.6%. If the conditional segmented incidence rate from years 5 to 7 is 5.5%, then the probability of new occurrence during the fifth to seventh year is approximately 91.6% multiplied by 5.5%, which is approximately 5.0%. Adding this new probability of occurrence to the five-year probability of HCC occurrence yields a seven-year probability of approximately 13.4%. Through this recursive method, the five-year probability of occurrence is always no less than the three-year probability, and the seven-year probability of occurrence is always no less than the five-year probability. Furthermore, assuming the segmented incidence rates pre-meet the effective probability range, the cumulative probability of occurrence will not exceed one.

[0077] When the overall risk score is lower than the lowest score anchor point in the segmented incidence rate conversion table for patients with sustained virological response, the probability conversion unit can use the segmented incidence rate corresponding to the lowest score anchor point for conversion. When the overall risk score is higher than the highest score anchor point, it can use the segmented incidence rate corresponding to the highest score anchor point for conversion and can output a prompt indicating that the score exceeds the anchor point range. If the conditional segmented incidence rate for a certain time period is missing in the conversion table, the probability conversion unit can stop outputting the multi-time-window incidence probability and prompt that the conversion table configuration is incomplete. Finally, the probability conversion unit outputs the three-year, five-year, and seven-year hepatocellular carcinoma incidence probabilities as multi-time-window incidence probabilities to the risk output unit. In this way, the probability conversion unit can convert the overall risk score into a recursive multi-time-window incidence probability that conforms to the long-term follow-up logic of treated chronic hepatitis B patients, so that the probability results have both temporal continuity and can reflect the risk accumulation process of patients at different follow-up stages.

[0078] Risk output unit 106 is used to generate risk stratification results based on the occurrence probability of multiple time windows and preset stratification thresholds.

[0079] In this embodiment, the risk output unit 106 receives the multi-time-window occurrence probabilities output by the probability conversion unit 105 and generates risk stratification results based on a preset stratification threshold. The risk stratification result refers to the system's graded output of the patient's current risk level for hepatocellular carcinoma based on the three-year, five-year, and seven-year occurrence probabilities. The risk stratification result can include low risk, medium risk, and high risk, and can be expanded to more levels such as extremely low risk, low risk, medium risk, high risk, and extremely high risk, depending on the follow-up management needs of medical institutions. The preset stratification threshold refers to a probability boundary value pre-set and stored before the system is put into use, used to determine which risk level the multi-time-window occurrence probabilities fall into. This threshold can be determined based on previous follow-up data of treated chronic hepatitis B patients, medical institution screening strategies, expert consensus, or clinical management guidelines, and remains fixed within the same version of the system to ensure consistent stratification results under the same input conditions.

[0080] The risk output unit 106 can stratify the incidence probabilities at three, five, and seven years separately, or it can perform comprehensive stratification based on the results of one main time window or a combination of multiple time windows. In one implementation, the system uses the five-year incidence probability of hepatocellular carcinoma as the primary stratification criterion, and the three-year and seven-year incidence probabilities as auxiliary display information. For example, if the five-year incidence probability is below 5%, the risk output unit 106 classifies the patient as low-risk; if the five-year incidence probability is between 5% and below 12%, the patient is classified as medium-risk; and if the five-year incidence probability is not less than 12%, the patient is classified as high-risk. If the three-year incidence probability has reached the high-risk threshold, even if the five-year incidence probability is within the medium-risk range, the risk output unit 106 can still raise the final risk stratification by one level according to preset rules to indicate that the patient has a significant short-term risk. If the seven-year incidence probability is high but the three-year and five-year incidence probabilities have not reached the high-risk threshold, the risk output unit 106 can maintain the current stratification while generating a long-term risk concern marker. The above thresholds and raising rules are only one implementation example; in actual applications, they can be set according to the target population data.

[0081] For example, if a patient's probability of developing hepatocellular carcinoma (HCC) is 2.5% in three years, 6.8% in five years, and 10.5% in seven years, the risk output unit 106 can classify this patient as medium risk based on a preset stratification threshold and output that the patient needs to be managed according to a medium-risk follow-up strategy. As another example, if another patient's probability of developing HCC is 7% in three years, 13% in five years, and 20% in seven years, the risk output unit 106 can classify this patient as high risk. Furthermore, if a patient's probability of developing HCC is 6% in three years but 10% in five years, falling within the medium-risk range, the system can output a high-risk or medium-high-risk alert based on short-term risk adjustment rules, ensuring that the risk stratification results reflect situations where short-term events are more likely to occur.

[0082] The output of the risk output unit 106 may include patient identification information or de-identified ID, prediction date, comprehensive risk score, three-year probability of hepatocellular carcinoma, five-year probability of hepatocellular carcinoma, seven-year probability of hepatocellular carcinoma, final risk stratification result, and stratification basis. The stratification basis may include the main time window used, the corresponding probability of occurrence, the triggered stratification threshold, and whether there are short-term risk upsizing or long-term risk concern indicators. To facilitate clinical understanding, the risk output unit 106 can also provide the output results to doctors or health management systems in the form of tables, reports, electronic medical record prompts, follow-up management lists, or interface data. For example, the display interface may indicate "The current five-year probability of occurrence is 6.8%, falling into the medium-risk range, and the final stratification is medium-risk," and simultaneously display the three-year and seven-year probabilities of occurrence, but this prompt does not replace the doctor's clinical diagnosis and treatment decisions.

[0083] To ensure the traceability of the output results, the risk output unit 106 can store the input probability, the version of the stratification threshold used, the version of the stratification rule, the output time, and the operator information for each risk stratification process as a risk prediction record. When a new target patient dataset is generated during a subsequent patient follow-up, the risk output unit 106 can also compare the new risk stratification results with historical risk stratification results to generate risk change indicators. For example, if a patient changes from low risk to medium risk, the system can mark it as an increased risk; if it changes from high risk to medium risk, the system can mark it as a decreased risk; if the patient remains at medium or high risk multiple times consecutively, the system can mark it as a persistent risk state. Through the above recording method, doctors can view the risk evolution of patients at different follow-up points, thus making it easier to understand the relationship between the probability of occurrence of multiple time windows and the risk stratification results.

[0084] In cases of data anomalies or probability conversion failures, the risk output unit 106 may not generate explicit low, medium, or high risk conclusions, but instead output a prompt indicating insufficient data or the need for verification. For example, when the probability conversion unit 105 fails to provide the occurrence probabilities for three, five, and seven years, or when the occurrence probabilities significantly exceed the reasonable range of zero to one hundred percent, the risk output unit 106 may prevent the generation of risk stratification results and prompt for checking the comprehensive risk score, probability conversion relationship, or original test data. Through the above settings, the risk output unit 106 can convert the occurrence probabilities of multiple time windows into explicit, interpretable, storable, and comparable risk stratification results, enabling the prediction results of this system to be used for long-term risk management and individualized follow-up arrangements for treated chronic hepatitis B patients.

[0085] To further validate the risk prediction effectiveness of this system, a retrospective follow-up cohort of 1286 patients with chronic hepatitis B who received nucleoside or nucleotide analogue therapy and achieved sustained virological response for more than six months was selected for a validation study. All patients were not diagnosed with hepatocellular carcinoma at baseline and had complete data including age, sex, quantitative hepatitis B surface antigen (HBsAg) levels, albumin levels, aspartate aminotransferase (AST) levels, upper limit of normal for AST, and platelet count. 900 of these patients were used as the modeling cohort, and the remaining 386 were used as the internal validation cohort, with no overlap between the two. The median follow-up time for all patients was 6.4 years, ranging from 1.2 to 9.8 years. During the follow-up period, 94 cases of hepatocellular carcinoma occurred, including 66 in the modeling cohort and 28 in the internal validation cohort. The imbalance correction unit uses the preset correction table in the aforementioned implementation method. Specifically, the correction coefficient is 1.05 when the hepatitis B surface antigen risk level is one level higher than the APRI risk level, and 1.10 when it is two levels higher; the correction coefficient is 1.08 when the APRI risk level is one level higher than the hepatitis B surface antigen risk level, and 1.15 when it is two levels higher. A scoring method without a virus-fibrosis imbalance correction coefficient was used as an ablation control. The baseline risk score and comprehensive risk score were calculated using the same target patient dataset, and their predictive performance for the risk of hepatocellular carcinoma at three, five, and seven years was compared.

[0086] In the internal validation cohort, without the virus-fibrin imbalance correction factor, the predicted C-index for 3, 5, and 7 years were 0.742, 0.756, and 0.751, with 95% confidence intervals of 0.670–0.814, 0.691–0.821, and 0.687–0.815, respectively. With the addition of the virus-fibrin imbalance correction factor, the predicted C-index for 3, 5, and 7 years improved to 0.781, 0.803, and 0.795, with 95% confidence intervals of 0.713–0.849, 0.744–0.862, and 0.735–0.855, respectively. Using a bootstrap method with repeated sampling, the p-values ​​for improvement in the 3, 5, and 7 years C-index were 0.038, 0.021, and 0.029, respectively. Correspondingly, without the virus-fibrosis imbalance correction factor, the time-dependent AUCs for the 3-year, 5-year, and 7-year predictions were 0.755, 0.768, and 0.762, respectively, with 95% confidence intervals of 0.681–0.829, 0.702–0.834, and 0.697–0.827, respectively. After incorporating the virus-fibrosis imbalance correction factor, the time-dependent AUCs for the 3-year, 5-year, and 7-year predictions increased to 0.792, 0.817, and 0.806, respectively, with 95% confidence intervals of 0.723–0.861, 0.760–0.874, and 0.746–0.866, respectively, and corresponding p-values ​​of 0.044, 0.018, and 0.027, respectively. The evaluation of the 7-year time window was completed using a time-dependent evaluation method considering censored data, and the internal validation cohort had follow-up records of up to 9.8 years, thus enabling the validation of the risk of occurrence in 7 years.

[0087] In the calibration evaluation, the five-year incidence rate of hepatocellular carcinoma was the primary evaluation criterion. Without the virus-fibrosis imbalance correction factor, the calibration slope was 0.84 and the Brier score was 0.066. After adding the virus-fibrosis imbalance correction factor, the calibration slope increased to 0.96 and the Brier score decreased to 0.052, indicating an improved consistency between the predicted probability and the actual observed risk. When further stratified according to the five-year predicted risk, the scoring method without the virus-fibrosis imbalance correction coefficient divided the 386 cases into a low-risk group (n=141), a medium-risk group (n=151), and a high-risk group (n=94). Within five years, these three groups had 3, 10, and 15 cases of hepatocellular carcinoma, respectively, corresponding to actual incidence rates of 2.1%, 6.6%, and 16.0%. After incorporating the virus-fibrosis imbalance correction coefficient, the 386 cases were divided into a low-risk group (n=148), a medium-risk group (n=149), and a high-risk group (n=89). Within five years, these three groups had 2, 10, and 16 cases of hepatocellular carcinoma, respectively, corresponding to actual incidence rates of 1.4%, 6.7%, and 18.0%. Therefore, after incorporating the virus-fibrosis imbalance correction coefficient, some patients who actually developed hepatocellular carcinoma were reclassified into a higher-risk stratum. The actual event rate decreased in the low-risk group and increased in the high-risk group, making the gradient between risk strata clearer. Further comparison of the average predicted probability and the actual incidence rate of each group revealed that, after correction, the average predicted probabilities of the low-risk group, medium-risk group, and high-risk group were 1.6%, 7.2%, and 18.2%, respectively, which were closer to the corresponding actual incidence rates of 1.4%, 6.7%, and 18.0%, indicating that the stratification results and probability outputs have good calibration consistency.

[0088] Simultaneously, this system was compared with PAGE-B, mPAGE-B, and CAMPAS risk scores. During the comparison, PAGE-B, mPAGE-B, and CAMPAS were all directly applied to the same internal validation cohort according to their original scoring rules, without refitting parameters in this cohort, to evaluate the direct applicability of existing scores in patients who had received treatment and achieved sustained virological response. In the internal validation cohort, the C-indexes of PAGE-B at three, five, and seven-year time windows were 0.701, 0.718, and 0.713, respectively; those of mPAGE-B were 0.718, 0.734, and 0.728; and those of CAMPAS were 0.731, 0.746, and 0.739. The C-indexes of this system, after adding a virus-fibrosis imbalance correction coefficient, were 0.781, 0.803, and 0.795, all higher than the aforementioned existing risk scores. Using CAMPAS as a better-performing control among existing scoring methods, our system showed a C-index improvement of 0.036 over a five-year time window compared to CAMPAS. Reclassification evaluation showed that, compared to the scoring method without a virus-fibrinosis imbalance correction coefficient, our system achieved a net reclassification improvement index of 0.146 over a five-year time window (95% confidence interval: 0.035–0.257, P-value: 0.011), and a composite discriminant improvement index of 0.038 (95% confidence interval: 0.011–0.065, P-value: 0.006). Compared to CAMPAS, our system achieved a net reclassification improvement index of 0.214 over a five-year time window (95% confidence interval: 0.076–0.352, P-value: 0.004), and a composite discriminant improvement index of 0.052 (95% confidence interval: 0.019–0.085, P-value: 0.002). The above validation results indicate that in chronic hepatitis B patients who have achieved sustained virological response after nucleoside or nucleotide analog treatment, converting the risk level difference between the hepatitis B surface antigen risk parameter and the APRI risk parameter into a virus-fibrosis imbalance correction coefficient, and using this correction coefficient to correct the baseline risk score, can improve the discriminative ability, calibration consistency, and risk reclassification effect of hepatocellular carcinoma risk prediction. This supports the use of this system for individualized patient follow-up management and determination of hepatocellular carcinoma screening intensity.

[0089] Although this application discloses preferred embodiments as described above, it is not intended to limit this application. Any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of this application. Therefore, the scope of protection of this application should be determined by the scope defined in the claims of this application.

Claims

1. A risk prediction system for hepatocellular carcinoma in patients with chronic hepatitis B, characterized in that, include: The data acquisition unit is used to acquire the age, sex, quantitative value of hepatitis B surface antigen, albumin value, aspartate aminotransferase value, upper limit of normal value of aspartate aminotransferase, and platelet count of chronic hepatitis B patients who have achieved sustained virological response after nucleoside or nucleotide analog treatment, forming a target patient dataset. The parameter generation unit is used to generate hepatitis B surface antigen risk parameters based on the quantitative value of hepatitis B surface antigen in the target patient dataset, and to calculate APRI value and APRI risk parameters based on aspartate aminotransferase value, upper limit of normal aspartate aminotransferase value and platelet count. The imbalance correction unit is used to compare the risk level difference between the hepatitis B surface antigen risk parameter and the APRI risk parameter, and to generate a virus-fibrosis imbalance correction coefficient when the risk level corresponding to the hepatitis B surface antigen risk parameter is inconsistent with the risk level corresponding to the APRI risk parameter. The scoring unit is used to generate a baseline risk score based on age, gender, albumin level, hepatitis B surface antigen risk parameter, and APRI risk parameter, and then corrects the baseline risk score based on the virus-fibrosis imbalance correction coefficient to obtain a comprehensive risk score. The probability conversion unit is used to generate multiple time window probabilities based on the comprehensive risk score, including the probability of occurrence of hepatocellular carcinoma in three years, five years, and seven years. The risk output unit is used to generate risk stratification results based on the occurrence probability of multiple time windows and preset stratification thresholds.

2. The hepatocellular carcinoma risk prediction system for patients with chronic hepatitis B according to claim 1, characterized in that, The imbalance correction unit is specifically used for: Read the quantitative value of hepatitis B surface antigen, the risk level of hepatitis B surface antigen and its corresponding classification interval from the hepatitis B surface antigen risk parameters, and read the APRI value, APRI risk level and its corresponding classification interval from the APRI risk parameters; The risk levels of hepatitis B surface antigen and APRI are converted into level codes, and an initial directed imbalance value is generated based on the difference between the two level codes. The viral side interval position value is generated based on the position of the quantitative value of hepatitis B surface antigen in its respective grade interval, and the fibrosis side interval position value is generated based on the position of the APRI value in its respective grade interval. When the initial directed imbalance value indicates that the risk level of hepatitis B surface antigen is higher than the risk level of APRI, the virus-side interval position value and the fibrosis-side interval position value are used to determine whether the initial directed imbalance value belongs to the virus-dominant type of credible imbalance; when the initial directed imbalance value indicates that the risk level of APRI is higher than the risk level of hepatitis B surface antigen, the fibrosis-side interval position value and the virus-side interval position value are used to determine whether the initial directed imbalance value belongs to the fibrosis-dominant type of credible imbalance. When the credible imbalance is either virologically dominant or fibroticly dominant, a virological-fibrotic imbalance correction coefficient is generated based on the direction and absolute value of the initial directed imbalance value. When the credible imbalance is not virologically dominant, the absolute value of the initial directed imbalance value is reduced by one level to generate the virological-fibrotic imbalance correction coefficient.

3. The hepatocellular carcinoma risk prediction system for patients with chronic hepatitis B according to claim 1, characterized in that, The parameter generation unit is specifically used for: Read the preset reference threshold package for sustained virological response population, which includes the 33rd percentile threshold, 67th percentile threshold, upper reference boundary of the highest risk interval and lower reference boundary of the lowest risk interval for the quantitative value of hepatitis B surface antigen, as well as the 33rd percentile threshold, 67th percentile threshold, upper reference boundary of the highest risk interval and lower reference boundary of the lowest risk interval for the APRI value; The detection boundary processing is performed on the quantitative values ​​of hepatitis B surface antigen in the target patient dataset to obtain the effective quantitative value of hepatitis B surface antigen. Based on the comparison results between the effective quantitative value of hepatitis B surface antigen and the 33rd percentile threshold and the 67th percentile threshold of the quantitative value of hepatitis B surface antigen in the reference threshold package for the sustained virological response population, the risk level of hepatitis B surface antigen, the level code of hepatitis B surface antigen and the classification interval to which hepatitis B surface antigen belongs are generated. The APRI value is calculated based on the aspartate aminotransferase (AST) value, the upper limit of normal aspartate aminotransferase (AST) value, and the platelet count. Based on the comparison between the APRI value and the 33rd percentile threshold and the 67th percentile threshold of the APRI value in the reference threshold package for sustained virological response population, the APRI risk level, APRI level code, and the APRI classification interval are generated. When the effective hepatitis B surface antigen quantitative value or APRI value is in the corresponding highest risk interval or lowest risk interval, the upper reference boundary of the highest risk interval or the lower reference boundary of the lowest risk interval in the reference threshold package for the sustained virological response population is called to supplement the grade interval to which the hepatitis B surface antigen belongs or the grade interval to which the APRI belongs, so that it has a definite upper and lower boundary. The effective quantitative value of hepatitis B surface antigen, the risk level of hepatitis B surface antigen, the grade code of hepatitis B surface antigen, and the grade interval to which hepatitis B surface antigen belongs are packaged into hepatitis B surface antigen risk parameters, and the APRI value, APRI risk level, APRI grade code, and the grade interval to which APRI belongs are packaged into APRI risk parameters.

4. The hepatocellular carcinoma risk prediction system for patients with chronic hepatitis B according to claim 3, characterized in that, The scoring unit is specifically used for: Read the preset scoring table, and generate age score, gender score, albumin score, hepatitis B surface antigen score and APRI score respectively based on age, gender, albumin value, hepatitis B surface antigen risk parameter and APRI risk parameter, and accumulate them to generate a basic risk score; The imbalance correction pathway is determined based on the relationship between the hepatitis B surface antigen (HBsAg) grade code and the APRI grade code. When the HBsAg grade code is greater than the APRI grade code, it is determined to be the viral residual correction pathway. When the APRI grade code is greater than the HBsAg grade code, it is determined to be the fibrosis pressure correction pathway. When the two are the same, the imbalance correction pathway is closed. Liver reserve gating coefficients were generated based on albumin scores. When albumin scores were 0, 1, and 2, the liver reserve gating coefficients for the viral residual correction channel were 1.00, 0.90, and 0.80, respectively, and the liver reserve gating coefficients for the fibrosis pressure correction channel were 1.00, 1.10, and 1.20, respectively. When the viral residual correction channel is activated, the difference between the hepatitis B surface antigen score and the virus-fibrosis imbalance correction coefficient minus 1 is multiplied and then multiplied by the liver reserve gating coefficient to generate the viral residual increment score, while the fibrosis pressure increment score is set to 0. When the fibrosis pressure correction channel is open, the APRI score is multiplied by the difference between the virus-fibrosis imbalance correction coefficient and the value after subtracting 1, and then multiplied by the liver reserve gating coefficient to generate the fibrosis pressure increment score, while the virus residual increment score is set to 0; when the imbalance correction channel is closed, both the virus residual increment score and the fibrosis pressure increment score are set to 0. The comprehensive risk score is obtained by adding the basic risk score, the viral residual increment score, and the fibrotic stress increment score.

5. The hepatocellular carcinoma risk prediction system for patients with chronic hepatitis B according to claim 1, characterized in that, The probability conversion unit is specifically used for: Read the preset conversion table of the incidence rate of the sustained virological response population. The conversion table of the incidence rate of the sustained virological response population includes multiple comprehensive risk score anchor points, and the conditional segmentation incidence rate of 0 to 3 years, conditional segmentation incidence rate of 3 to 5 years and conditional segmentation incidence rate of 5 to 7 years corresponding to each comprehensive risk score anchor point. Among them, the conditional segmentation incidence rate of the same time period does not decrease as the comprehensive risk score anchor point increases, and the cumulative incidence probability obtained by recursively extrapolating the remaining non-occurrence probability does not exceed 1. Based on the comprehensive risk score, adjacent low-score anchor points and adjacent high-score anchor points are determined in the segmented incidence conversion table of the sustained virological response population, and the score position value of the comprehensive risk score between the adjacent low-score anchor points and adjacent high-score anchor points is generated. Based on the score position value, the 0-3 year condition segment incidence rate, 3-5 year condition segment incidence rate, and 5-7 year condition segment incidence rate corresponding to adjacent low score anchor points and adjacent high score anchor points are interpolated in order-preserving manner to generate the 0-3 year condition segment incidence rate, 3-5 year condition segment incidence rate, and 5-7 year condition segment incidence rate of the current patient. Each condition segment incidence rate obtained by interpolation is located between the corresponding condition segment incidence rates of adjacent low score anchor points and adjacent high score anchor points. The incidence rate of conditional segments from 0 to 3 years is used as the probability of hepatocellular carcinoma in three years. The probability of hepatocellular carcinoma in five years is generated by adding the product of the three-year probability of occurrence, the three-year probability of non-occurrence, and the conditional segmentation of the 3- to 5-year incidence rates. The probability of hepatocellular carcinoma in five years is generated by adding the product of the probability of no occurrence in five years and the conditional segmentation of occurrence from 5 to 7 years. The probability of hepatocellular carcinoma in seven years is then generated by adding the probability of hepatocellular carcinoma in three years, five years, and seven years as the multi-time-window occurrence probabilities.

Images