A hepatitis b population screening method based on viral load and immune response characteristics

Abstract

The application discloses a hepatitis B population screening method based on viral load and immune response characteristics, relates to the field of medical data analysis, and comprises collecting viral load detection data and liver biochemical index detection data in a preset period, and respectively constructing a viral load sequence and a liver biochemical index sequence in time sequence; cross-correlation analysis is performed on the two sequences, and a time shift corresponding to the maximum cross-correlation function value is taken as a first lag time; frequency domain transformation is performed on the two sequences, a phase under a main frequency is obtained, and a phase difference is calculated; the main frequency period is converted into a second lag time; the first lag time and the second lag time are compared with a first threshold value, a second threshold value, a third threshold value and a fourth threshold value, and an immune response sluggish type or an immune response active type is marked and a screening priority is adjusted, so that the problem that screening stratification is unstable due to the difficulty in quantifying the misplacement of viruses and immune responses and the cycle difference is solved; the application solves the problem that screening stratification is unstable due to the difficulty in quantifying the misplacement of viruses and immune responses and the cycle difference in the hepatitis B population.

Description

Technical Field

[0001] This invention relates to the field of medical data analysis, and more specifically, to a method for screening hepatitis B patients based on viral load and immune response characteristics. Background Technology

[0002] Hepatitis B is a chronic infectious disease primarily characterized by hepatitis B virus (HBV) infection. The interaction between viral replication and the body's immune response determines the degree of hepatocellular damage, the level of inflammatory activity, and the rate of subsequent fibrosis progression. In clinical follow-up management, commonly used observational information includes viral load test results and liver biochemical indicators reflecting liver function and inflammation levels, such as serum alanine aminotransferase (ALT), aspartate aminotransferase (AST), gamma-glutamyl transferase (GGT), and albumin. These test data naturally possess time-series attributes, reflecting the dynamic process of viral replication and liver inflammation changes in patients over a period of time.

[0003] In existing screening and follow-up practices, a common challenge lies in the significant differences in the strength, timing, and rhythm of immune responses among different patients. This can lead to inconsistencies or even misalignments between changes in viral load and fluctuations in liver biochemical indicators. Some patients experience rapid increases in biochemical indicators after changes in viral load, while others show significant delays or only slight fluctuations. Simultaneously, some patients exhibit cyclical inflammatory fluctuations, which are difficult to capture with single or infrequent testing. In population screening scenarios, relying solely on the level of a single test or the absolute level of a single indicator can easily lead to two problems: firstly, individuals with weak immune responses and consistently stable biochemical indicators may be classified as low-risk, delaying crucial follow-up; secondly, individuals with strong immune responses resulting in short-term increases in biochemical indicators may receive excessive attention, leading to resource misallocation. Due to the lack of quantitative descriptions that can characterize the temporal and rhythmic relationships between viral load sequences and liver biochemical marker sequences, the screening system struggles to stably distinguish different immune response patterns and perform stratified management in large populations, thus affecting the efficiency of continuous follow-up and early risk identification in hepatitis B patients. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a method for screening hepatitis B patients based on viral load and immune response characteristics, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: A method for screening hepatitis B patients based on viral load and immune response characteristics, comprising: Collect viral load detection data and liver biochemical index detection data of the target population within a preset period, construct a viral load sequence in chronological order from the viral load detection data, and construct a liver biochemical index sequence in chronological order from the liver biochemical index detection data. Cross-correlation analysis was performed on the viral load sequence and the liver biochemical index sequence to obtain cross-correlation function values ​​at different time shifts, and the time shift with the largest corresponding cross-correlation function value was determined as the first lag time. The viral load sequence and the liver biochemical index sequence are subjected to frequency domain transformation to obtain their phase at the main frequency, and the phase difference is calculated. The phase difference is converted into a time difference according to the period corresponding to the main frequency, which is used as the second lag time to characterize the difference in periodic immune response. The following groups in the target population are marked as having a sluggish immune response and given higher screening priority: those whose first lag time is greater than a preset first threshold and whose second lag time is greater than a preset second threshold. The following groups within the target population are marked as having an active immune response and have their screening priority reduced: those whose first lag time is less than the third threshold and whose second lag time is less than the fourth threshold; The first threshold is greater than the third threshold, and the second threshold is greater than the fourth threshold.

[0006] Preferably, the liver biochemical index detection data includes at least serum alanine aminotransferase (ALT) index, or further includes one or more of the following indexes: aspartate aminotransferase (AST) index, gamma-glutamyl transferase (γ-GT) index, and albumin index. When multiple indicators are included, the indicators are normalized and weighted to obtain the liver biochemical indicator detection data.

[0007] Preferably, constructing the viral load sequence and the liver biochemical marker sequence includes: Viral load detection data and liver biochemical index detection data are collected at preset time intervals within the preset period. When the actual collection time is not completely consistent with the preset time interval, time interpolation is performed to make up the detection data between adjacent collection time points.

[0008] Preferably, performing cross-correlation analysis on the viral load sequence and the liver biochemical marker sequence includes: The viral load sequence and the liver biochemical index sequence were subjected to amplitude normalization. Within the preset lag time search range, calculate the corresponding cross-correlation function value for each candidate lag time; Within the lag time search range, the candidate lag time with the largest cross-correlation function value is selected as the first lag time.

[0009] Preferably, frequency domain transformation of the viral load sequence and the liver biochemical index sequence includes: The viral load sequence and the liver biochemical index sequence were subjected to discrete Fourier transforms to obtain their respective spectra; The non-zero frequency component with the largest amplitude in each spectrum is selected as the main frequency, and the phase of the main frequency component of viral load and the phase of the main frequency component of liver biochemical indicators are obtained accordingly. The phase difference between the phase of the main frequency component of the viral load and the phase of the main frequency component of the liver biochemical index is taken as the phase difference, and converted into the second lag time according to the period corresponding to the main frequency.

[0010] Preferably, the first threshold and the second threshold are obtained through a training dataset, which includes viral load detection data and liver biochemical index detection data for each subject, and the subjects in the training dataset are manually divided into immune-sluggish type and immune-active type.

[0011] Preferably, the method for obtaining the first threshold, the second threshold, the third threshold, and the fourth threshold specifically includes: calculating the corresponding first lag time and the second lag time for each subject in the training dataset; Within a preset candidate threshold search space, different first, second, third, and fourth thresholds are combined. For threshold combinations, subjects in the training dataset are automatically categorized according to the criteria for determining sluggish and active immune responses. For each set of threshold combinations, a consistency index is calculated between the automatic classification result and the manual classification result. The consistency index is the ratio of the number of subjects whose automatic classification result is the same as the manual classification result to the total number of subjects in the training dataset. Select the threshold combination that maximizes the consistency index from the candidate threshold search space.

[0012] Preferably, before obtaining different threshold combinations, a lag feature vector is constructed for each subject with a first lag time as the first component and a second lag time as the second component, and cluster analysis is performed on the set of lag feature vectors to obtain multiple clusters; For each cluster, a first ratio is calculated based on the total number of subjects within the cluster that are manually labeled as having a sluggish immune response, and the cluster with the largest first ratio is determined as the first candidate cluster. For each cluster, a second ratio is calculated based on the total number of subjects within the cluster that are manually labeled as having an active immune response, and the cluster with the largest second ratio is determined as the second candidate cluster. The first candidate cluster is used to determine the first threshold and the second threshold. The second candidate cluster is used to determine the third threshold and the fourth threshold.

[0013] Preferably, the first threshold is obtained by subtracting N standard deviations from the average of the first lag times of all subjects in the first candidate cluster; the second threshold is obtained by subtracting N standard deviations from the average of the second lag times of all subjects in the first candidate cluster; the third threshold is obtained by adding N standard deviations to the average of the first lag times of all subjects in the second candidate cluster; and the fourth threshold is obtained by adding N standard deviations to the average of the second lag times of all subjects in the first candidate cluster.

[0014] Preferably, the method further includes: Trajectory feature vectors were extracted from the viral load sequence and liver biochemical index sequence of each subject in the target population. The trajectory feature vectors include the first lag time, the second lag time, the maximum value of the cross-correlation function, the peak time of the viral load sequence, and the peak time of the liver biochemical index sequence. Cluster analysis was performed on the trajectory feature vector set of the target population to obtain multiple trajectory type clusters, and each subject was labeled with the trajectory type label corresponding to its trajectory type cluster. For historical subjects with subsequent follow-up outcome annotations, the ratio of the number of historical subjects with progressive outcomes within each trajectory type cluster to the total number of historical subjects within that trajectory type cluster was calculated. Trajectory type clusters with ratios greater than a preset progress ratio threshold are identified as high-progression trajectory type clusters, and trajectory type clusters with ratios less than a preset stability ratio threshold are identified as stable trajectory type clusters. When a subject in the target population is labeled as belonging to the high-progression trajectory type cluster, the screening priority of that subject is increased; when a subject in the target population is labeled as belonging to the stable trajectory type cluster, the screening priority of that subject is decreased.

[0015] The advantage of this invention over existing technologies lies in its addressing the temporal misalignment between viral load changes and liver biochemical index fluctuations in the hepatitis B population, and the difficulty in stable stratification due to periodic fluctuations in some individuals. This invention introduces a first lag time and a second lag time as immune response characteristics, enabling more accurate differentiation of different immune response patterns and priority management during large-scale screening. The first lag time is obtained through cross-correlation analysis. Its advantage lies in not relying on a single peak or individual time point, but rather comprehensively comparing the overall alignment of two time series within a preset lag range. Even under conditions of noise, local abnormal fluctuations, and inconsistencies in sampling, it can still identify the dominant delay relationship between viral load changes and liver biochemical index changes, thereby capturing the temporal misalignment characteristic of viral mutation preceding inflammatory response. The second lag time is obtained by converting the phase difference at the dominant frequency into a time difference. Its advantage lies in characterizing differences in periodic immune responses: when a patient's biochemical indicators show repetitive fluctuations or rhythmic variations, relying solely on time-domain correlation may be affected by amplitude changes and local peak-to-trough drift. However, the phase difference can reflect the sequential relationship and synchronicity of two sequences in the main periodic components, making it more suitable for identifying immune patterns such as rhythm asynchrony and periodic response lag. These two types of parameters capture different aspects of change from the overall delay relationship in the time domain and the periodic synchronization relationship in the frequency domain, respectively, forming a complementary relationship. This makes the identification of individuals with sluggish immune responses more stable, while avoiding misjudging short-term fluctuations caused by active immune responses as long-term high risk, thereby improving the accuracy of screening stratification and the efficiency of resource allocation.

[0016] This invention further provides a data organization method for multi-indicator scenarios, ensuring that liver biochemical indicator detection data includes at least serum alanine aminotransferase and can be combined with indicators such as aspartate aminotransferase, gamma-glutamyl transferase, and albumin. Through normalization and weighting, unified liver biochemical indicator detection data is constructed, enhancing adaptability to different clinical collection protocols. Addressing the issue of irregular sampling times in actual follow-up, this invention uses preset time intervals for collection and time interpolation to ensure that viral load sequences and liver biochemical indicator sequences can be analyzed on a unified time scale, improving the stability of cross-correlation and frequency domain analysis. For the cross-correlation and frequency domain calculation process, this invention provides specific implementation paths such as amplitude normalization, lag time search range, discrete Fourier transform, and determination of the dominant frequency and phase difference, reducing uncertainties in engineering implementation.

[0017] This invention introduces a training dataset and manually partitioned results. It determines threshold combinations through a combined search of the candidate threshold search space and maximizing the consistency index. Prior to this, cluster analysis is performed on the lag feature vector set to form candidate clusters. Thresholds are then constructed using the mean and standard deviation, improving the transferability and maintainability of threshold settings across different populations. Furthermore, the introduction of trajectory feature vectors and trajectory type clusters expands screening from a single lag judgment to trajectory type-driven hierarchical management. Combined with follow-up outcome annotation and statistical progress ratios, it enables priority screening of high-progression trajectory type clusters and downgraded screening of stable trajectory type clusters, further improving the efficiency and accuracy of large-scale population screening. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the overall process of the present invention; Figure 2 This is a schematic diagram of the immune typing logic for subjects based on multi-level thresholds. Figure 3 This is a schematic diagram of the first lag time determined based on cross-correlation analysis in this invention; Figure 4 This is a schematic diagram of the second lag time based on frequency domain transformation phase difference conversion in this invention; Figure 5 This is a schematic diagram of clustering based on trajectory feature vectors. Detailed Implementation

[0019] The specific embodiments of the present invention will now be described with reference to the accompanying drawings.

[0020] The core of this invention lies in utilizing the time-series relationship between viral load detection data and liver biochemical index detection data to characterize the temporal delay and circadian rhythm differences in the immune response. Based on this, the target population is then labeled as either immune-insensitive or immune-active, and screening priority management is implemented accordingly. Figure 1 As shown, the method forms a closed-loop process from data acquisition, sequence construction, hysteresis feature calculation, threshold determination to genotyping and priority adjustment. The purpose of this design is to avoid missed screening or over-screening caused by relying solely on a single test value, and to incorporate the more stable dynamic feature of immune response patterns into population stratification.

[0021] In one embodiment, viral load and liver biochemical marker data of the target population are collected within a preset period. The preset period can be set as a continuous time interval according to the organization of the screening task, and can range from 28 days to 365 days, with 56 days to 180 days being commonly used. Viral load data can be expressed using clinically common hepatitis B virus nucleic acid quantitative detection results, with units of IU / mL.

[0022] Liver biochemical marker detection data should include at least serum alanine aminotransferase (ALT), and may further include aspartate aminotransferase (AST), gamma-glutamyl transferase (γ-GT), and albumin to improve coverage of different inflammatory activities and liver synthetic function status. If multiple liver biochemical markers are collected, they can be normalized and weighted before being used as liver biochemical marker detection data to facilitate unified correlation analysis with viral load sequences. Normalization can use either min-max normalization or Z-score normalization. Min-max normalization maps the markers to the 0-1 range, facilitating the fusion of markers with different dimensions; Z-score normalization transforms the markers to a scale with a mean of 0 and a standard deviation of 1, reducing individual baseline differences. Weighting methods can use either preset fixed weights or data-driven weights. Preset fixed weights can be set according to the focus of the screening task; for example, the weights of ALT, AST, γ-GT, and albumin can be set to 0.4, 0.3, 0.2, and 0.1, respectively. Data-driven weights can be selected from the training dataset by aiming to make the fused sequence more consistent with the results of manual partitioning. The weight values ​​can be limited to 0 to 1 and the sum can be 1 to ensure that the fusion result is interpretable and reproducible.

[0023] In a further embodiment, when constructing the viral load sequence and liver biochemical index sequence, viral load detection data and liver biochemical index detection data are collected at preset time intervals. The preset time interval can be set to 1 to 30 days, commonly 7 to 14 days, to balance clinical sampling feasibility and time series analysis accuracy. When the actual collection time is not completely consistent with the preset time interval, time interpolation is performed to fill in the gaps between adjacent collection time points. The purpose of this setting is to align the sequences on a unified time axis and reduce the sensitivity of cross-correlation analysis and frequency domain transformation to unequal sampling intervals. The interpolation method can be any one of linear interpolation, spline interpolation, or piecewise conserved interpolation. Linear interpolation has low computational cost and is suitable for large-scale screening; spline interpolation can reconstruct curves more smoothly and is suitable for populations with significant index fluctuations; piecewise conserved interpolation can avoid introducing over-smoothing when sampling is sparse. The interpolated time sampling points can be set as a set of standard sampling points corresponding to the preset time interval.

[0024] In one embodiment, cross-correlation analysis is used to calculate the first lag time. For example... Figure 3As shown, cross-correlation analysis was performed on the viral load sequence and the liver biochemical index sequence to obtain cross-correlation function values ​​under different time shifts, and the time shift with the largest corresponding cross-correlation function value was determined as the first lag time. Cross-correlation analysis is used to capture the optimal alignment position of the overall trend of two sequences in time. Its advantage is that it does not rely on a certain peak point, but uses the whole sequence information for alignment, which is suitable for handling real follow-up data with noise, local abnormal fluctuations, or multi-peak structures. To improve computational stability, the viral load sequence and the liver biochemical index sequence can be normalized in amplitude, and then the cross-correlation function value can be calculated for each candidate lag time within a preset lag time search range. The candidate lag time with the largest cross-correlation function value within this search range is selected as the first lag time. The lag time search range can be set to 0 days to 120 days, and 0 days to 56 days is commonly used to cover the possible delay span between changes in viral load and the response of biochemical indicators. The cross-correlation function value can be calculated in the form of discrete cross-correlation. T The sequence, assuming the viral load sequence is... x(t) The liver biochemical index sequence is y(t) The candidate lag time is Then the cross-correlation function value It can be calculated using the following formula: ; in Indicates the first Normalized viral load values ​​at each time point Indicates the first Normalized values ​​of liver biochemical indicators at each time point This represents the time shift amount in units of time sampling intervals.

[0025] In another embodiment, the second lag time is obtained through frequency domain transformation and time conversion of the phase difference. For example... Figure 4 As shown, frequency domain transformation was performed on the viral load sequence and liver biochemical index sequence. Phase was extracted at the same dominant frequency, and the phase difference was calculated. The phase difference was then converted into a time difference according to the period corresponding to that dominant frequency, serving as the second lag time characterizing differences in periodic immune responses. The purpose of using the phase difference is to characterize the sequential relationship and synchronicity of the two sequences in their main periodic components. When liver biochemical indicators exhibit periodic fluctuations or repetitive oscillations, the phase difference can reflect the rhythmic lag or lead more stably than simple time domain correlation.

[0026] Frequency domain transformation can be performed using the Discrete Fourier Transform. Let the viral load sequence be... The liver biochemical index sequence is , respectively and Perform a discrete Fourier transform to obtain the spectrum. and In a specific embodiment, the viral load spectrum is... The frequency corresponding to the non-zero frequency component with the largest amplitude is determined as the dominant frequency. and at the main frequency Obtain from each location phase and phase Phase difference Represented as: ; The phase difference is calculated according to the period corresponding to the main frequency. Converted into time difference, the second lag time is obtained. The expression: ; in , This is the second lag time.

[0027] In one embodiment, the target population is immunophenotyped based on a first lag time and a second lag time, and the screening priority is adjusted accordingly. For example... Figure 2 As shown, the following groups within the target population are marked as having a sluggish immune response and given higher screening priority: the criteria are a first lag time greater than a preset first threshold and a second lag time greater than a preset second threshold. The following groups within the target population are marked as having an active immune response and given lower screening priority: the criteria are a first lag time less than a third threshold and a second lag time less than a fourth threshold, with the first threshold greater than the third threshold and the second threshold greater than the fourth threshold. The purpose of using dual thresholds to constrain the sluggish and active types is to avoid misjudgments based solely on a single lag characteristic, making the classification more robust. Screening priority can be implemented in several ways. Firstly, a screening task queuing method is used, adding subjects with a sluggish immune response to a high-priority queue and subjects with an active immune response to a low-priority queue, prioritizing retesting or further examinations for the high-priority queue under the same resource conditions. Secondly, a follow-up frequency adjustment method is used, setting shorter retesting intervals for subjects with a sluggish immune response and longer retesting intervals for subjects with an active immune response. Third, the triggering rule method sets a lower threshold for retesting in subjects with a sluggish immune response and a higher threshold for subjects with an active immune response, thus achieving differentiated resource allocation. The retesting interval can be set from 7 to 90 days, with 14 to 30 days being commonly used.

[0028] In another embodiment, the first and second thresholds are obtained through a training dataset. The training dataset includes viral load and liver biochemical marker data for each subject, and the subjects in the training dataset are manually categorized into those with a sluggish immune response and those with an active immune response. This manual categorization can be formed by clinicians based on long-term follow-up records, disease progression, and test results, or by multiple doctors independently labeling the data and then using a consistent decision-making process to form the final labels, thus reducing individual subjective bias. The training dataset can range from 100 to 100,000 participants, with 1,000 to 20,000 participants being commonly used.

[0029] In another embodiment, the method for obtaining the first, second, third, and fourth thresholds includes calculating the corresponding first and second lag times for each subject in the training dataset, and combining different first, second, third, and fourth thresholds within a preset candidate threshold search space. The candidate threshold search space can be constructed according to the quantile ranges of the first and second lag times in the training dataset to avoid meaningless extreme thresholds. The candidate threshold for the first lag time can be from 0 to 120 days, and the candidate threshold for the second lag time can be from 0 to 120 days; the third and fourth thresholds can be in lower ranges, such as from 0 to 60 days. For each threshold combination, the subjects in the training dataset are automatically genotyped according to the rules for determining a sluggish or active immune response, and a consistency index between the automatic genotyping result and the manual genotyping result is calculated. The consistency index is the ratio of the number of subjects whose automatic genotyping result is the same as the manual genotyping result to the total number of subjects in the training dataset. The threshold combination that maximizes the consistency index is selected from the candidate threshold search space. The purpose of this step is to make the typing as consistent as possible with human experience through interpretable threshold combinations without introducing complex models, so as to facilitate implementation in screening institutions.

[0030] In another embodiment, before obtaining different threshold combinations, a lag feature vector is constructed for each subject, with a first lag time as the first component and a second lag time as the second component. Cluster analysis is then performed on the set of lag feature vectors to obtain multiple clusters. The cluster analysis can employ any of K-means clustering, Gaussian mixture model clustering, or hierarchical clustering. K-means clustering is computationally efficient and suitable for large samples; Gaussian mixture model clustering can output the probability of cluster assignment and is suitable for non-spherical cluster shapes; hierarchical clustering facilitates the observation of merging relationships between clusters. For each cluster, a first ratio is calculated based on the total number of subjects manually labeled as having a sluggish immune response within the cluster, and the cluster with the largest first ratio is determined as the first candidate cluster. A second ratio is calculated based on the total number of subjects manually labeled as having an active immune response within the cluster, and the cluster with the largest second ratio is determined as the second candidate cluster. The first candidate cluster is used to determine the first and second thresholds, and the second candidate cluster is used to determine the third and fourth thresholds. The purpose of introducing clustering is to first find the subset of the population that is most consistent with the manual division in the feature space, so as to make the statistical derivation of the threshold more stable and reduce the threshold drift caused by boundary samples.

[0031] In another embodiment, thresholds are constructed based on statistics of the candidate clusters. The first threshold is calculated by subtracting N standard deviations from the average of the first lag times of all subjects within the first candidate cluster; the second threshold is calculated by subtracting N standard deviations from the average of the second lag times of all subjects within the first candidate cluster; the third threshold is calculated by adding N standard deviations to the average of the first lag times of all subjects within the second candidate cluster; and the fourth threshold is calculated by adding N standard deviations to the average of the second lag times of all subjects within the first candidate cluster. N can range from 0.5 to 3.0, but is commonly used from 1.0 to 2.0 to balance genotyping coverage and false positive rate. The purpose of using the mean and standard deviations to construct the thresholds is to make them adaptive to population distribution, facilitating reuse under data conditions in different regions and screening institutions.

[0032] In another embodiment, the present invention further introduces trajectory feature vectors and trajectory type clusters to provide trajectory type-level screening management in addition to immune typing. Figure 5As shown, trajectory feature vectors are extracted from the viral load sequence and liver biochemical index sequence of each subject in the target population. The trajectory feature vectors include the first lag time, the second lag time, the maximum value of the cross-correlation function, the peak time of the viral load sequence, and the peak time of the liver biochemical index sequence. The peak time can be determined by finding the time point corresponding to the maximum value in the sequence; when multiple peaks exist, the time point corresponding to the largest peak can be selected, or the peak that first exceeds a preset quantile threshold can be selected as the main peak. The quantile threshold can be set from 0.7 to 0.95. Cluster analysis is performed on the trajectory feature vector set to obtain multiple trajectory type clusters, and each subject is labeled with the trajectory type label corresponding to its respective trajectory type cluster. This trajectory type label is used to express the similarity of the subjects in the fluctuation patterns of viral load and liver biochemical indexes, making it easier for screening institutions to manage more individual differences with fewer categories.

[0033] In another embodiment, historical subjects with subsequent follow-up outcome annotations are used to calculate the ratio of the number of historical subjects with progressive outcomes within each trajectory type cluster to the total number of historical subjects within that trajectory type cluster. Trajectory type clusters with a ratio greater than a preset progression ratio threshold are identified as high-progression trajectory type clusters, and trajectory type clusters with a ratio less than a preset stability ratio threshold are identified as stable trajectory type clusters. The preset progression ratio threshold can be set to 0.2 to 0.8, and is commonly set to 0.4 to 0.6; the preset stability ratio threshold can be set to 0.05 to 0.5, and is commonly set to 0.1 to 0.3. Subjects belonging to high-progression trajectory type clusters in the target population are given higher screening priority, while subjects belonging to stable trajectory type clusters are given lower screening priority. This allows for priority coverage of individuals more likely to experience progression with the same screening resources, improving screening efficiency and management targeting.

[0034] The above embodiments illustrate the technical solution of the present invention. Adjustments to the data sampling frequency, interpolation method, normalization method, clustering algorithm, threshold search granularity, and threshold range under the same technical concept can all be implemented without departing from the inventive concept of the present invention.

[0035] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for screening hepatitis B patients based on viral load and immune response characteristics, characterized in that, include: Collect viral load and liver biochemical index data of the target population within a preset period, construct a viral load sequence in chronological order from the viral load data, and construct a liver biochemical index sequence in chronological order from the liver biochemical index data. Cross-correlation analysis was performed on the viral load sequence and the liver biochemical index sequence to obtain cross-correlation function values ​​at different time shifts, and the time shift with the largest corresponding cross-correlation function value was determined as the first lag time. The viral load sequence and the liver biochemical index sequence are subjected to frequency domain transformation to obtain their phase at the main frequency, and the phase difference is calculated. The phase difference is converted into a time difference according to the period corresponding to the main frequency, which is used as the second lag time to characterize the difference in periodic immune response. The following groups in the target population are marked as having a sluggish immune response and given higher screening priority: those whose first lag time is greater than a preset first threshold and whose second lag time is greater than a preset second threshold; The following groups within the target population are marked as having an active immune response and have their screening priority reduced: those whose first lag time is less than the third threshold and whose second lag time is less than the fourth threshold; The first threshold is greater than the third threshold, and the second threshold is greater than the fourth threshold.

2. The hepatitis B population screening method based on viral load and immune response characteristics according to claim 1, characterized in that, The liver biochemical index detection data includes at least the serum alanine aminotransferase (ALT) index, or may also include one or more of the following indicators: aspartate aminotransferase (AST) index, gamma-glutamyl transferase (γ-GT) index, and albumin index. When multiple indicators are included, the indicators are normalized and weighted to obtain the liver biochemical indicator detection data.

3. The hepatitis B population screening method based on viral load and immune response characteristics according to claim 1, characterized in that, Constructing the viral load sequence and the liver biochemical marker sequence includes: Viral load detection data and liver biochemical index detection data are collected at preset time intervals within the preset period. When the actual collection time is not completely consistent with the preset time interval, time interpolation is performed to make up the detection data between adjacent collection time points.

4. The hepatitis B population screening method based on viral load and immune response characteristics according to claim 1, characterized in that, The cross-correlation analysis of the viral load sequence and the liver biochemical marker sequence includes: The viral load sequence and the liver biochemical index sequence were subjected to amplitude normalization. Within the preset lag time search range, calculate the corresponding cross-correlation function value for each candidate lag time; Within the lag time search range, the candidate lag time with the largest cross-correlation function value is selected as the first lag time.

5. The hepatitis B population screening method based on viral load and immune response characteristics according to claim 1, characterized in that, Frequency domain transformation of the viral load sequence and the liver biochemical marker sequence includes: The viral load sequence and the liver biochemical index sequence were subjected to discrete Fourier transforms to obtain their respective spectra; The non-zero frequency component with the largest amplitude in each spectrum is selected as the main frequency, and the phase of the main frequency component of viral load and the phase of the main frequency component of liver biochemical indicators are obtained accordingly. The phase difference between the phase of the main frequency component of the viral load and the phase of the main frequency component of the liver biochemical index is taken as the phase difference, and converted into the second lag time according to the period corresponding to the main frequency.

6. The hepatitis B population screening method based on viral load and immune response characteristics according to claim 1, characterized in that, The first threshold and the second threshold are obtained through a training dataset, which includes viral load detection data and liver biochemical index detection data for each subject. The subjects in the training dataset are artificially divided into those with a sluggish immune response and those with an active immune response.

7. The hepatitis B population screening method based on viral load and immune response characteristics according to claim 6, characterized in that, The method for obtaining the first threshold, the second threshold, the third threshold, and the fourth threshold specifically includes: calculating the corresponding first lag time and the second lag time for each subject in the training dataset; Within a preset candidate threshold search space, different first, second, third, and fourth thresholds are combined. For threshold combinations, subjects in the training dataset are automatically categorized according to the criteria for determining sluggish and active immune responses. For each set of threshold combinations, a consistency index is calculated between the automatic classification result and the manual classification result. The consistency index is the ratio of the number of subjects whose automatic classification result is the same as the manual classification result to the total number of subjects in the training dataset. Select the threshold combination that maximizes the consistency index from the candidate threshold search space.

8. The hepatitis B population screening method based on viral load and immune response characteristics according to claim 7, characterized in that, Before obtaining different threshold combinations, a lag feature vector is constructed for each subject with the first lag time as the first component and the second lag time as the second component. Cluster analysis is performed on the set of lag feature vectors to obtain multiple clusters. For each cluster, the first ratio of the total number of subjects marked as immunocompromised within the cluster according to the manual division is calculated, and the cluster with the largest first ratio is determined as the first cluster to be used; For each cluster, the second ratio of the total number of subjects marked as having an active immune response within the cluster is calculated based on the manual division. The cluster with the largest second ratio is determined as the second candidate cluster. The first candidate cluster is used to determine the first threshold and the second threshold. The second candidate cluster is used to determine the third threshold and the fourth threshold.

9. The hepatitis B population screening method based on viral load and immune response characteristics according to claim 8, characterized in that: The first threshold is obtained by subtracting N standard deviations from the average of the first lag times of all subjects in the first candidate cluster; the second threshold is obtained by subtracting N standard deviations from the average of the second lag times of all subjects in the first candidate cluster; the third threshold is obtained by adding N standard deviations to the average of the first lag times of all subjects in the second candidate cluster; and the fourth threshold is obtained by adding N standard deviations to the average of the second lag times of all subjects in the first candidate cluster.

10. The hepatitis B population screening method based on viral load and immune response characteristics according to claim 1, characterized in that, The method further includes: Trajectory feature vectors were extracted from the viral load sequence and liver biochemical index sequence of each subject in the target population. The trajectory feature vectors include the first lag time, the second lag time, the maximum value of the cross-correlation function, the peak time of the viral load sequence, and the peak time of the liver biochemical index sequence. Cluster analysis was performed on the trajectory feature vector set of the target population to obtain multiple trajectory type clusters, and each subject was labeled with the trajectory type label corresponding to its own trajectory type cluster; For historical subjects with subsequent follow-up outcome annotations, the ratio of the number of historical subjects with progressive outcomes within each trajectory type cluster to the total number of historical subjects within that trajectory type cluster was calculated. Trajectory type clusters with ratios greater than a preset progress ratio threshold are identified as high-progression trajectory type clusters, and trajectory type clusters with ratios less than a preset stability ratio threshold are identified as stable trajectory type clusters. When a subject in the target population is labeled as belonging to the high-progression trajectory type cluster, the screening priority of that subject is increased; when a subject in the target population is labeled as belonging to the stable trajectory type cluster, the screening priority of that subject is decreased.

Images