Whole course management system based on fusion analysis of hepatitis b serological and virology indexes

Abstract

The present application relates to the technical field of medical data mining, in particular to a whole-course management system based on fusion analysis of hepatitis B serological and virological indexes, which comprises a serum index collection module, an evolution path construction module, a characteristic section extraction module, an intervention response classification module and a fusion structure induction module.In the present application, based on the index sequence matrix formed by continuous sampling, the trend trajectory is generated by tracking the change direction of the index concentration at each time node, the change fragments with structural differences are extracted in combination with the biochemical response markers, the index combination relationship under different interventions is classified and arranged, the matching mapping between the intervention behavior and the response path is established, the regular characteristics of the intervention effect are reflected through the combination of multiple indexes, and then the induction analysis result set covering the whole treatment process is generated, thereby providing clear data basis for dynamic intervention evaluation and treatment path adjustment.

Description

Technical Field

[0001] This invention relates to the field of medical data mining technology, and in particular to a full-course management system based on the fusion analysis of hepatitis B serological and virological indicators. Background Technology

[0002] The field of medical data mining technology primarily involves the analysis and processing of medical data to extract information with practical application value. This field includes, but is not limited to, the processing and analysis of medical information such as medical records, diagnostic results, treatment plans, and clinical pathways. Core aspects encompass data mining algorithms, machine learning techniques, health data analysis, and disease prediction models. The goal of medical data mining is to discover potential patterns and trends from large amounts of health data, optimize processes such as disease diagnosis, treatment plan development, and health management, and improve the efficiency and quality of medical services. With the rapid development of big data technology and artificial intelligence, medical data mining has gradually become one of the important tools in the modern medical industry, widely applied in personalized medicine, precision medicine, disease prediction, chronic disease management, and many other areas. Among these, the traditional full-course management system based on the fusion analysis of hepatitis B serological and virological indicators refers to establishing a comprehensive, end-to-end management system by integrating the serological and virological data of hepatitis B patients. This system aims to monitor and analyze hepatitis B-related indicators in patients in real time and adjust treatment strategies promptly based on changes in serological and virological data. Traditional hepatitis B treatment management methods often rely on single data indicators or manual experience, lacking systematic and comprehensive analysis, resulting in an inability to fully assess the patient's treatment effectiveness and disease progression. This patent provides a management tool that continuously tracks the entire treatment process of patients by integrating multiple medical indicators for data analysis and intelligent decision-making, which helps to optimize treatment plans and improve patient adherence.

[0003] Current technologies lack structured capture of the changing trends of serological and virological indicators across multiple time points during hepatitis B treatment. Static analysis of indicator values ​​alone is insufficient to identify the dynamic evolutionary relationships between indicators. In clinical practice, there is a lack of systematic classification and quantitative analysis of the indicator linkages triggered by treatment interventions. This results in a lack of clear trajectory references for intervention effect feedback, and the treatment pathway lacks response support at the indicator level. This can easily lead to problems such as delayed response, untimely adjustments, and biased judgment of patient status during the treatment cycle. Summary of the Invention

[0004] To address the technical problems existing in the prior art, embodiments of the present invention provide a full-course management system based on the fusion analysis of hepatitis B serological and virological indicators. On one hand, a full-course management system based on the fusion analysis of hepatitis B serological and virological indicators is provided, the system comprising: The serum index collection module is equipped with a blood collection device to collect blood samples at multiple time points during the treatment of hepatitis B patients, extract indicators of hepatitis B surface antigen, hepatitis B virus and alanine aminotransferase, and construct a timeline-aligned serum index sequence matrix. The evolution path construction module analyzes the concentration change direction of hepatitis B surface antigen and hepatitis B virus in the serum index sequence matrix, marks the concentration trend path, records the response direction of alanine aminotransferase at the trend path node, and generates trajectory path structure data. The feature segment extraction module filters trajectory segments with a combination of increasing, stable and turning features in the trajectory path structure data, performs differential labeling, and generates biological change segment data groups; The intervention response classification module compares the consistency between the biological change segment data group and the indicator trajectory combination and segment change label under the intervention operation, removes the intervention sequences with differential expression, and generates the intervention impact relationship table data structure. The fusion structure summarization module statistically analyzes the number of decreasing fragments of hepatitis B surface antigen, the proportion of stable fragments of hepatitis B virus, and the number of fluctuations of alanine aminotransferase in the data structure of the intervention influence relationship table, and constructs a set of summary analysis results for the entire treatment course.

[0005] As a further aspect of the present invention, the serum indicator sequence matrix includes hepatitis B surface antigen concentration sequence, hepatitis B virus concentration sequence, and alanine aminotransferase concentration sequence; the trajectory path structure data includes concentration change direction annotation, path node response information, and multi-indicator joint trend type; the biological change segment data group includes increasing characteristic segments, stable characteristic segments, and turning characteristic segments; the intervention influence relationship table data structure includes intervention behavior type, indicator trajectory combination form, and segment change label matching relationship; and the full-course inductive analysis result set includes the number of surface antigen decreasing segments, the proportion of stable viral segments, and enzyme fluctuation frequency characteristics.

[0006] As a further aspect of the present invention, the intervention sequence for eliminating differential expression refers to the exclusion of indicator trajectory combinations that show inconsistencies with biological change segments under intervention conditions.

[0007] As a further aspect of the present invention, the labeled concentration trend path refers to tracking and recording the development trend based on the direction of change in serum index concentration.

[0008] As a further aspect of the present invention, the serum index collection module includes: The data acquisition submodule acquires raw blood samples from multiple treatment time points in the blood collection device, labels the samples according to patient number and time label, sorts the samples according to time sequence, and generates a time series identifier mapping matrix. The indicator extraction submodule, based on the time series identifier mapping matrix, calls the numbered samples to extract the values ​​of hepatitis B surface antigen, hepatitis B virus, and alanine aminotransferase, and after classification and recording, obtains a multi-indicator synchronous sampling data table. The sequence construction submodule aligns similar indicators according to time nodes based on the multi-indicator synchronous sampling data table, integrates the number and indicator dimension, establishes a two-dimensional mapping relationship between indicators and time, and obtains a serum indicator sequence matrix.

[0009] As a further aspect of the present invention, the evolution path construction module includes: The concentration trend determination submodule, based on the serum index sequence matrix, calls the hepatitis B surface antigen concentration value and hepatitis B virus concentration value at consecutive time nodes, calculates the difference between adjacent time nodes, and classifies the increase or decrease according to the positive or negative direction of the difference to obtain the concentration change direction sequence. The trend path labeling submodule assigns path numbers to the markers of consecutive time nodes based on the concentration change direction sequence, merges consecutive nodes with the same change direction into a path segment, and binds the change direction with the path segment marker to obtain multiple trend path label sequences. The response trajectory construction submodule calls the multi-segment trend path label sequence to obtain the alanine aminotransferase values ​​of the corresponding nodes, and marks them according to the direction of change between nodes. The marking results are bidirectionally merged with the trend path sequence to establish a multi-index sequence that corresponds to the joint change in concentration and the response direction, and generate trajectory path structure data.

[0010] As a further aspect of the present invention, the feature segment extraction module includes: Based on the trajectory path structure data, the trajectory filtering submodule detects the corresponding concentration change tags in all continuous trajectory segments, filters the set of trajectory segments that simultaneously include three change modes: concentration increase, concentration stability, and concentration transition, and adds a structure index and boundary position to each trajectory segment to obtain a composite change segment sequence. The change recognition submodule calls the composite change segment sequence, extracts the index change combination of time nodes in each trajectory segment, calculates the duration and frequency of each type of change combination in the trajectory segment, and filters and classifies according to the set combination weight interval to obtain the index combination feature label set. The difference labeling submodule compares the feature label sequences between adjacent trajectory segments based on the feature label set of the index combination, calculates the label offset rate and sets the feature offset threshold, marks the segment group with the difference degree greater than the offset threshold, adds an identification number, and generates a biological change segment data group.

[0011] As a further aspect of the present invention, the intervention response classification module includes: The intervention matching submodule obtains the combination of indicator trajectories for each intervention sequence under the corresponding time period based on the biological change segment data group, calculates the consistency ratio between the trajectory combination and the segment change label, sorts them according to consistency, numbers and classifies them, and establishes an intervention consistency distribution matrix. The differential screening submodule extracts the intervention sequence numbers whose consistency ratio is lower than the intervention response deviation threshold based on the intervention consistency distribution matrix, removes the sequences from the full intervention set, updates the sequence index table and the identifier list, and obtains the high-difference intervention exclusion set; The corresponding table construction submodule calls the remaining intervention sequences outside the high-difference intervention exclusion set, pairs them one-to-one with the corresponding indicator trajectory combinations, constructs mapping entries between intervention behaviors and indicator change paths, and generates a structured two-dimensional record table to obtain the intervention impact relationship table data structure.

[0012] As a further aspect of the present invention, the fusion structure summarization module includes: Based on the data structure of the intervention impact relationship table, the indicator feature statistics submodule extracts the indicator path corresponding to each intervention behavior, and calculates the number of all concentration-decreasing fragments in hepatitis B surface antigen, the ratio of the number of static and stable fragments in hepatitis B virus to the total number of fragments, and the number of fluctuations in alanine aminotransferase, to obtain the indicator change feature parameter set. The combined feature construction submodule calls the indicator change feature parameter set, combines the three types of parameter values ​​corresponding to the intervention behavior as independent feature vector entries, constructs a corresponding table based on the intervention behavior number, marks the numerical range and label number under each feature combination, and obtains the intervention feature combination coding table. The treatment course summary generation submodule extracts the combination feature entries of all intervention behaviors within the entire treatment course based on the intervention feature combination coding table, groups them by coding labels and accumulates the number, and establishes a structured set by combining the grouping statistical results to generate a set of full treatment course summary analysis results.

[0013] As a further aspect of the present invention, the intervention response deviation threshold is a numerical limit for determining whether the consistency between the intervention sequence and the biological change segment deviates.

[0014] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: Based on the index sequence matrix formed by continuous sampling, trend trajectories are generated by tracking the changes in index concentrations at each time point. Combined with biochemical response markers, change segments with structural differences are extracted, and the relationship between index combinations under different interventions is classified and organized. A matching mapping between intervention behaviors and response paths is established, and the regular characteristics of intervention effects are reflected through the combined performance of multiple indicators. In turn, an inductive analysis result set covering the entire treatment process is generated, providing clear data basis for dynamic intervention evaluation and treatment path adjustment. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a schematic diagram of the system of the present invention; Figure 2 This is a flowchart of the serum index collection module in this invention; Figure 3 This is a flowchart of the evolution path construction module in this invention; Figure 4 This is a flowchart of the feature segment extraction module in this invention; Figure 5 This is a flowchart of the intervention response classification module in this invention; Figure 6 This is a flowchart of the fusion structure summarization module in this invention. Detailed Implementation

[0017] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0018] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0019] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0020] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0021] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0022] This invention provides a full-course management system based on the fusion analysis of hepatitis B serological and virological indicators, such as... Figure 1 The diagram shown illustrates a full-course management system based on the fusion analysis of hepatitis B serological and virological indicators. This system includes: The serum marker acquisition module is equipped with a blood collection device to continuously sample multiple time points during the treatment of hepatitis B patients, and performs the extraction of markers such as hepatitis B surface antigen, hepatitis B virus and alanine aminotransferase, and constructs a timeline-aligned serum marker sequence matrix. The evolutionary path construction module is based on the serum index sequence matrix. It analyzes the concentration change direction of hepatitis B surface antigen and hepatitis B virus at continuous time points, marks the trend path of concentration increase and decrease, records the response direction of alanine aminotransferase at the path node, and outputs trajectory path structure data including the joint change trend of multiple indicators. The feature segment extraction module, based on trajectory path structure data, filters trajectory segments in continuous segments that exhibit a combination of three indicators: increasing concentration, stable concentration, and concentration inflection point. It then performs differential labeling on all trajectory segments to generate a biological change segment data set. The intervention response classification module compares the consistency between the combination of indicator trajectories and the segment change labels under each intervention operation based on the biological change segment data group, filters out intervention sequences with high differential expression, organizes the corresponding table of intervention behavior and indicator change path, and generates the intervention impact relationship table data structure. The fusion structure summarization module, based on the data structure of the intervention impact relationship table, judges the combined performance characteristics of the number of decreasing fragments of hepatitis B surface antigen, the proportion of stable and continuous fragments of hepatitis B virus, and the number of fluctuations in alanine aminotransferase under the intervention behavior, and constructs a set of summarization analysis results for the entire treatment course.

[0023] The serum indicator sequence matrix includes the concentration sequences of hepatitis B surface antigen, hepatitis B virus, and alanine aminotransferase. The trajectory path structure data includes concentration change direction annotations, path node response information, and multi-indicator joint trend types. The biological change segment data group includes increasing characteristic segments, stable characteristic segments, and turning point characteristic segments. The intervention influence relationship table data structure includes intervention behavior types, indicator trajectory combination forms, and segment change label matching relationships. The full course of treatment summary analysis result set includes the number of surface antigen decreasing segments, the proportion of stable viral segments, and the characteristics of enzyme fluctuation frequency.

[0024] Specifically, such as Figure 2 As shown, the serum marker acquisition module includes: The data acquisition submodule acquires raw blood samples from multiple treatment time points in the blood collection device, labels the samples according to patient number and time label, sorts the samples according to time sequence, and generates a time series identifier mapping matrix. First, based on the collection information recorded in the blood collection device, the patient number, collection device number, timestamp, sample number, and sample type are read from each sampling record. The collection time is standardized and converted into timestamp data accurate to the second. This timestamp is then paired and bound to the corresponding patient number. By traversing all the original record data, a one-to-many mapping relationship is established for multiple sample timestamps corresponding to the same patient number. During the mapping process, all timestamps for each patient are sorted in ascending order and assigned time sequence numbers sequentially, such as the first time point being sequence number 1, the second being sequence number 2, and so on, forming sample sequence labels. At the same time, the correspondence between the original collection device number and this label is recorded. Subsequently, the patient number, sample timestamp, and sequence number are... The sequence label, sample timestamp, and time sequence label are combined into a set of mapping data to form an initial time series identifier mapping matrix. During the data anomaly screening process, the time interval between two adjacent time points is compared. If the difference between adjacent timestamps is less than the preset minimum sampling interval, it is considered as duplicate or abnormal collection and needs to be removed and re-labeled. In practical applications, the minimum sampling interval is set to be no less than 3 hours. That is, if the difference between two time tags is less than 10,800 seconds, the system determines it as invalid data and removes it. For samples collected by the device in continuous mode, adjacent samples are merged based on the number of samples collected by different devices, and duplicate records are further screened by the sampling batch number. After all processing is completed, a mapping matrix containing patient number, sample timestamp, and time sequence label is output.

[0025] The indicator extraction submodule uses the time series identifier mapping matrix to call numbered samples, extract the values ​​of hepatitis B surface antigen, hepatitis B virus, and alanine aminotransferase, and obtain a multi-indicator synchronous sampling data table after classification and recording. Based on the time-series identifier mapping matrix, the unique sample identifier composed of each patient number and time label is retrieved one by one. The system retrieves the blood test records corresponding to that time point from the sample data analysis module, and calls up various serum biological indicator data in the records. In this process, the system not only extracts hepatitis B surface antigen, hepatitis B virus load, and alanine aminotransferase, but also needs to expand the extraction scope to complete HBV serological markers, specifically including hepatitis B surface antigen (HBsAg), hepatitis B surface antibody (Anti-HBs), hepatitis B e antigen (HBeAg), hepatitis B e antibody (Anti-HBe), hepatitis B core antibody (Anti-HBc), and hepatitis B core antibody IgM (Anti-HBc). IgM, commonly known as the "Hepatitis B two and a half pairs" full-item index, is extracted along with serum biochemical indicators. In addition to ALT, aspartate aminotransferase (AST) values ​​also need to be extracted simultaneously. Furthermore, to improve the diagnostic and treatment monitoring system, the system also needs to include alpha-fetoprotein (AFP) and relevant results from abdominal ultrasound screening for hepatocellular carcinoma (HCC). During the extraction process, for immune indicators such as hepatitis B surface antigen, the system retrieves the corresponding test item name and reads the concentration value. For hepatitis B viral load, it identifies numerical entries containing the keyword HBV-DNA. For ALT and AST, the corresponding enzyme activity values ​​are extracted from the indicator names. For AFP and abdominal ultrasound results, structured extraction is performed based on the AFP concentration value and the imaging conclusions in the ultrasound report. Simultaneously, the difference between the detection time and the collection time of each indicator needs to be compared. If the detection time of any indicator is earlier than or later than the sampling time by more than [a certain amount], [further action will be taken]. If the time interval is 20 minutes, the value is marked as invalid and will not be written to the system. At the same time, it is necessary to determine whether each indicator is missing or below the measurable limit. If there are cases where the measurement result is "negative", "unmeasurable", or blank, the data must also be marked as empty and the reason for the missing data must be recorded. During the process of summarizing valid data, all the above-mentioned expanded indicator values ​​are synchronously merged into a single record and aligned with the time-order labels to establish a multi-indicator data table indexed by patient number and sequence label. For example, for the third sample corresponding to a certain patient number, if both the core indicator value and the AST value exist, the record is written to an array containing all values. If the ALT or AST value is greater than the set abnormal threshold (e.g., ALT greater than 50, AST greater than 40), the system automatically adds an asterisk to it to indicate that the value is too high. Finally, the system completes the extraction and recording of the indicator values ​​corresponding to all samples and obtains a multi-indicator synchronous sampling data table.

[0026] The sequence construction submodule aligns similar indicators by time nodes based on the multi-indicator synchronous sampling data table, integrates the number and indicator dimensions, establishes a two-dimensional mapping relationship between indicators and time, and obtains the serum indicator sequence matrix. First, the system extracts the indicator records for each patient at all time points. All indicators of the same category are arranged chronologically, forming hepatitis B surface antigen (HBsAg) sequences, hepatitis B virus (HBV) sequences, alanine aminotransferase (ALT) sequences, aspartate aminotransferase (AST) sequences, alpha-fetoprotein (AFP) sequences, and other relevant serological marker sequences. Within each sequence category, the data are arranged chronologically to form a univariate time series vector. For missing time point data, the system does not perform interpolation but directly marks the position as null. For consecutively missing start or end data, if the indicator has a stable mean range across all samples, the average value of the indicator at the same sequence position in other patients is allowed as supplementation. During the construction process, the system simultaneously records the indicators... Whether the values ​​are original samples or supplementary fillers is determined for subsequent processing. After the indicator sequences are generated, multiple indicator sequences of the same patient, including ALT, AST, HBsAg, HBV-DNA, AFP, etc., are integrated into a set of structure records as their indicator sequence information. At the same time, a two-dimensional structure is established in the matrix space with the time sequence label as the horizontal axis and all extracted biochemical and immune indicators as the vertical axis. The patient number, time label, and indicator dimension are bound and mapped to construct a serum indicator sequence matrix. Each row of the matrix represents the value of multiple indicators at a time node, and each column shows the change of a certain indicator at different time nodes, thus completely constructing an indicator sequence matrix suitable for subsequent sequence processing.

[0027] Specifically, such as Figure 3 As shown, the evolutionary path construction module includes: The concentration trend determination submodule is based on the serum index sequence matrix. It calls the hepatitis B surface antigen concentration value and hepatitis B virus concentration value at continuous time nodes, calculates the difference between adjacent time nodes, and classifies the increase or decrease according to the positive or negative direction of the difference to obtain the concentration change direction sequence. The system sequentially extracts the hepatitis B surface antigen (HBsAg) concentration sequence and hepatitis B virus (HBV) concentration sequence for each patient at all time points, and numbers and organizes the sequences chronologically. Then, for each indicator, the system sequentially compares the concentration values ​​between two consecutive time points, subtracting the concentration value of the previous time point from the current time point's concentration value. The calculated difference is used to determine the direction of concentration change. This process is repeated for each pair of adjacent time points, generating a continuous difference sequence. In the difference judgment stage, a positive difference is recorded as "increasing," a negative difference as "decreasing," and a zero difference as "remaining unchanged," forming a continuous sequence of change direction markers. To ensure the effectiveness of the direction judgment, amplitude filtering is performed on each group of differences, setting a minimum judgment threshold. For HBsAg concentration, the threshold is set to 0.1 IU / mL, meaning that when the absolute value of the difference is less than 0.1, it is marked as "remaining unchanged," regardless of whether it is positive or negative. For HBV concentration, the threshold is set to 100. The change is recorded as "increase" or "decrease" only if the number of copies / mL exceeds this threshold range. For newly included AFP indicators, corresponding clinically significant change thresholds (such as 10 ng / mL) are also set for similar judgment. The system continuously updates the directional sequence throughout the judgment process to ensure that the complete change trend of each indicator sequence is clearly marked, and finally obtains the concentration change directional sequence.

[0028] The trend path labeling submodule assigns path numbers to the markers of consecutive time nodes based on the concentration change direction sequence, groups consecutive nodes with the same change direction into a path segment, and binds the change direction with the path segment marker to obtain multiple trend path label sequences. The system scans and identifies the direction of change for patients across all time points. Starting from the first time point, the direction of change at the current point is read as the starting direction and set as the starting point of the current path segment. The system then checks whether the direction of change of the next node is consistent with the direction of the current path segment. If they are consistent, the current path segment continues to expand; otherwise, the current path segment is terminated, and the starting node number, ending node number, and corresponding direction of change are recorded. The path segments are numbered sequentially according to their appearance. The next direction of change is then used as the starting point of a new path segment, and the above operation is repeated until all time points have been traversed. During the formation of path segments, the system records the length, start and end positions, and direction of each path segment. If a path segment contains only one node, it is not recorded as a valid path segment. Only combinations of two or more consecutive nodes with the same direction can constitute a path segment. After all paths are labeled, the path segments for hepatitis B surface antigen, viral concentration, and other key indicators such as possible AFP are recorded as independent trend path label sequences, providing a structural basis for the next step of indicator response analysis, ultimately resulting in multiple trend path label sequences.

[0029] The response trajectory construction submodule calls multiple trend path label sequences to obtain the alanine aminotransferase values ​​of the corresponding nodes, and marks them according to the direction of change between nodes. The marking results are bidirectionally merged with the trend path sequences to establish a multi-index sequence that corresponds to the joint change in concentration and the response direction, and generate trajectory path structure data. For each labeled trend path segment, its corresponding time node range is located. Then, the ALT and AST values ​​corresponding to each node within the segment are extracted from the serum indicator sequence matrix. The system reads the ALT and AST values ​​sequentially according to the time order within the trend path segment, forming a biochemical response subsequence for that segment. Subsequently, the values ​​of adjacent nodes in the subsequence are compared, and the difference between each pair of nodes is calculated. If the ALT difference between adjacent nodes is positive and greater than 5 U / L, the change in that segment is marked as "increasing"; if the difference is negative and less than -5 U / L, it is marked as "decreasing"; if the difference is within ±5 U / L, it is marked as "decreasing". Changes within the U / L range are recorded as "unchanged". Similarly, for the AST index, the system sets corresponding thresholds based on its clinical fluctuation characteristics for synchronous judgment, marking it as "rising", "falling" or "unchanged" for AST. The system forms the corresponding biochemical response direction based on the overall trend of ALT and AST changes in each path segment. If multiple difference directions within a path segment are inconsistent, the direction with the largest proportion is taken as the main direction response label for that segment. Then, the ALT and AST response directions are compared with the concentration change directions of the original trend path segment to determine whether they are synergistic changes. If the biochemical indicator change direction is the same as or consistent with the concentration trend direction, it is recorded as "synergistic". If they are completely inconsistent, it is marked as "non-synergistic". Finally, the ALT and AST response directions are bidirectionally bound to the trend path label of each segment to establish a joint recording sequence covering multiple virological and biochemical enzymatic indicators on the same path, and output a structured trajectory path data structure.

[0030] Specifically, such as Figure 4 As shown, the feature segment extraction module includes: The trajectory filtering submodule detects the concentration change labels corresponding to all continuous trajectory segments based on trajectory path structure data, filters the trajectory segment set that includes three change modes: concentration increase, concentration stability and concentration transition, and adds a structure index and boundary position to each trajectory segment to obtain the composite change segment sequence. First, all path segment information for each patient is retrieved. The concentration change direction labels within each path segment are used as the initial screening criteria. Each path segment is examined to see if it forms a continuous segment in chronological order, and all concentration change direction labels within that continuous segment are extracted. Then, the path segments are classified and statistically analyzed according to label type. If a continuous segment simultaneously contains labels of "increasing," "stable," and "turning point," it is identified as a composite change trajectory segment. In actual judgment, the system's determination of the "increasing" label depends on the proportion of the preceding indicator value showing an upward trend. If more than 60% of the nodes within a path segment show a continuous increase, it is marked as increasing. For the "stable" label... The judgment is based on the indicator change amplitude being lower than the preset fluctuation range. The judgment range is set as the absolute value of the indicator difference being less than 0.1 units. The identification of the "turning point" label depends on the number of direction changes. If there is more than one direction change from rising to falling or from falling to rising within the path segment, the segment is marked as a turning point. After the system performs the above screening on all trajectory segments, it extracts the segments that meet the three conditions into composite trajectory segments and assigns a unique structural index number to each segment. This number is generated according to the patient number and the starting position of the time. At the same time, the starting node position and the ending node position of the segment in the sequence are recorded, and finally, a composite change segment sequence is obtained.

[0031] The change recognition submodule calls the composite change segment sequence, extracts the combination of indicator changes at time nodes in each trajectory segment, calculates the duration and frequency of each type of change combination in the trajectory segment, and filters and classifies according to the set combination weight interval to obtain the indicator combination feature label set. The system iterates through the time nodes within each composite segment, extracting the combination of concentration change directions for all indicators at each node. Specifically, for each time node, the system acquires the change direction markers for indicators such as hepatitis B surface antigen, hepatitis B virus, alanine aminotransferase, and aspartate aminotransferase, and combines three or more of them to form change combination labels, such as "rise-fall-rise-rise" or "stable-stable-fall-level". Then, the frequency of this combination appearing throughout the entire path is statistically analyzed, and the system is divided into sub-segments according to temporal continuity, recording the start and end nodes of each combination and its duration. The number of nodes is used as the duration. Then, the weights of all appearing combination labels are calculated, with the weight range set from 0 to 1. The system calculates the frequency of the combination in the path segment based on the frequency divided by the total number of nodes. If the proportion is between 0.3 and 0.6, it is considered a moderate feature; if it exceeds 0.6, it is marked as a primary feature; and if it is below 0.3, it is a secondary feature. The weight range can be manually adjusted according to the sample size. The system selects combination labels with a weight greater than 0.4 and records their corresponding feature information based on these combinations. Finally, the system categorizes the combination labels that meet the conditions in each trajectory segment into the indicator combination feature label set.

[0032] The difference labeling submodule compares the feature label sequences between adjacent trajectory segments based on the indicator combination feature label set, calculates the label offset rate and sets the feature offset threshold, marks the segment group with the difference greater than the offset threshold, adds an identification number, and generates a biological change segment data group. First, the system extracts the combined label types and their weight values ​​of adjacent trajectory segments and establishes a label sequence comparison matrix. By comparing the label names one by one and their weight differences, the system calculates the offset rate between the two label sets. This offset rate is defined as the average value of the label weight differences. During the comparison process, the system calculates the corresponding values ​​for labels with the same name and assigns a value of 1 to inconsistent labels as the maximum offset. The system summarizes the offset values ​​of all labels to calculate the average offset rate. The feature offset threshold is set to 0.35. If the calculated average offset rate is greater than this threshold, the two trajectory segments are considered to have significant differences. The system marks the segment group as a difference segment and generates a unique identifier number for it. This number is based on the previous trajectory segment number with the addition of a difference suffix. Finally, all segments marked as differences are output as biological change segment data groups.

[0033] Specifically, such as Figure 5 As shown, the intervention response classification module includes: The intervention matching submodule obtains the combination of indicator trajectories for each intervention sequence within the corresponding time period based on the biological change segment data set, calculates the consistency ratio between the trajectory combination and the segment change label, sorts them by consistency, numbers and classifies them, and establishes an intervention consistency distribution matrix. First, each intervention sequence entry is traversed, the corresponding time period range is extracted, and the set of nodes in the serum index change path that overlap with the time period is called. The combination of index change directions of all time nodes within the time period is extracted and used as the index trajectory combination form associated with the intervention sequence. Then, the system extracts the feature label set of each segment from the biological change segment data group, compares each intervention trajectory combination with the segment label one by one, records the number of matching labels and the total number of labels, and calculates the label consistency ratio between the two. This ratio is obtained by dividing the number of matching labels by the total number of labels. After the system has calculated the ratio of all intervention sequences, these ratios are sorted in descending order, arranged from high to low. Each intervention sequence is assigned a unique number and classified. If multiple intervention sequences have the same ratio, the order of intervention start time is used as the second sorting condition. After sorting, a record table is constructed that associates the intervention sequence number with the corresponding consistency ratio, forming an intervention consistency distribution matrix.

[0034] The differential screening submodule extracts the intervention sequence numbers whose consistency ratio is lower than the intervention response deviation threshold based on the intervention consistency distribution matrix, removes the sequences from the full intervention set, updates the sequence index table and the identifier list, and obtains the high-difference intervention exclusion set; First, the system extracts the intervention sequence number and its corresponding consistency ratio from each row. Then, it performs a filtering operation on each ratio, setting the intervention response deviation threshold to 0.4. That is, when the ratio of an intervention sequence to a certain biological change segment is less than 0.4, the system determines that the intervention sequence deviates from that segment and records the intervention sequence number as a potential removal target. Subsequently, the system aggregates the set of all intervention sequence numbers with ratios below the threshold, removes duplicates, and generates a preliminary exclusion sequence set. Each number in this set is further checked to see if it has low consistency records in multiple segments. If an intervention sequence is below the threshold in more than two segments, it is officially added to the high-difference removal list. After removal, these numbers are removed from the original intervention sequence set, and the sequence index table is updated to renumber all remaining intervention sequences to ensure continuity. The entries in the intervention sequence identifier list are also updated synchronously. Finally, the system outputs a set of intervention sequences that do not contain high-difference interventions as the high-difference intervention exclusion set.

[0035] The corresponding table construction submodule calls the remaining intervention sequences outside the high-difference intervention exclusion set, pairs them one-to-one with the corresponding indicator trajectory combinations, constructs the mapping entries between intervention behavior and indicator change path, and generates a structured two-dimensional record table to obtain the intervention impact relationship table data structure; The system reads the time range and intervention type identifier of each intervention sequence one by one, and searches for indicator change path segments that overlap with its time range in the serum indicator change path database. When matching, the time node number is used as the comparison basis. If the start and end numbers of several pre-time periods are the same as any path segment, a one-to-one pairing relationship is established, and the intervention behavior number is bound to the indicator change path segment. Then, the system writes each pairing relationship into a structured two-dimensional record table. The table uses the intervention number as the row index and the indicator path number as the column index. The cells record the indicator combination label information and the response direction information of ALT and AST corresponding to the pairing. All pairing relationships are written into the table in sequence to form a complete set of entries corresponding to the intervention impact. At the same time, the system performs a structural check on the table to ensure that each intervention behavior is bound to only one indicator path. If there is time overlap or path conflict, the system records the conflict information but does not write it into the formal table. Finally, the record table is output as an intervention impact relationship table data structure.

[0036] Specifically, such as Figure 6 As shown, the fusion structure summarization module includes: The indicator feature statistics submodule is based on the data structure of the intervention impact relationship table. It extracts the indicator path corresponding to each intervention behavior, and calculates the number of all concentration-decreasing fragments in hepatitis B surface antigen, the ratio of the number of static and stable fragments in hepatitis B virus to the total number of fragments, and the number of fluctuations in alanine aminotransferase, to obtain the indicator change feature parameter set. First, the system iterates through each intervention record, extracts the corresponding indicator path number for each intervention, and retrieves the complete sequence of three or more indicator changes based on the path number. For the hepatitis B surface antigen (HBsAg) indicator sequence, the system sequentially reads the concentration values ​​of each adjacent time node and performs a difference judgment operation. If the value of the later node is less than that of the previous node and the difference meets the threshold requirement, it is determined to be a decreasing segment and the number is accumulated. For the hepatitis B virus (HBV) indicator, the system determines the number of occurrences of continuous stable segments and calculates its ratio to the total number of segments. Then, the system processes the change sequences of alanine aminotransferase (ALT) and aspartate aminotransferase (AST). The system compares the values ​​of each node pair, counts the sum of the number of upward and downward fluctuations, and records them as ALT and AST fluctuation count parameters. At the same time, the system checks whether the AFP test value in this time period exceeds the abnormal threshold (such as 20 ng / mL or 400 ng / mL) and whether there is a space-occupying lesion record in the abdominal ultrasound. If there is an abnormality, the number of risk markers is recorded. Finally, the values ​​of the above multiple indicator feature parameters are recorded in the form of structured fields and output as a set of indicator change feature parameters.

[0037] The combined feature construction submodule calls the indicator change feature parameter set, combines the three types of parameter values ​​corresponding to the intervention behavior as independent feature vector entries, constructs a corresponding table based on the intervention behavior number, marks the numerical range and label number under each feature combination, and obtains the intervention feature combination coding table. Using the intervention behavior number as an index, the corresponding feature parameter values ​​are extracted one by one, namely the number of hepatitis B surface antigen concentration decrease fragments, the hepatitis B virus quiescent-to-stable ratio, the number of ALT and AST fluctuations, and AFP and ultrasound abnormality markers. These values ​​are combined to form feature vector entries, and each feature vector is recorded in the form of a structure. At the same time, the system assigns a unique label number to each combined vector. The label number is classified according to the numerical range. For example, the HBsAg decrease number is classified into levels A, B, and C, the HBV quiescent ratio is classified into levels X, Y, and Z, and the number of ALT and AST fluctuations is classified into levels alpha, beta, and gamma. The system names each feature vector according to its level combination. For example, if the decrease level is B, the quiescent level is Y, and the fluctuation level is beta, then the combined label number is "BY-beta". This label number is mapped to the intervention number and written into the corresponding table. The table records the intervention number, feature vector value, corresponding range level, and label number. Finally, the feature combination coding of all intervention behaviors is completed, and the intervention feature combination coding table is constructed.

[0038] The treatment course summary generation submodule extracts the combination feature items of all intervention behaviors within the entire treatment course based on the intervention feature combination coding table, groups them by coding labels and accumulates the number, and establishes a structured set based on the grouping statistical results to generate the full treatment course summary analysis result set; First, all intervention numbers appearing throughout the entire treatment course are summarized and grouped according to the combination feature labels established in the table. All intervention behaviors with the same label are grouped together, and the number of intervention behaviors in each group is counted. The group number, number of members, and composition details are recorded. Then, all combination labels are sorted in ascending order to create a structured results table. Each row corresponds to a label combination, and each column records the list of intervention numbers, the frequency of the combination in the entire treatment course, and the proportion of the total number of intervention behaviors. The proportion is obtained by dividing the number of interventions in each group by the total number of interventions. The system is then reordered from high to low frequency to form a complete statistical distribution table of intervention combinations. Based on this, the system establishes a structured set structure, assigns a unique record entry to each label combination, and summarizes the results into a summary analysis set of the entire treatment course.

[0039] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A full-course management system based on the fusion analysis of hepatitis B serological and virological indicators, characterized in that the system... include: The serum index collection module is equipped with a blood collection device to collect blood samples at multiple time points during the treatment of hepatitis B patients, extract indicators of hepatitis B surface antigen, hepatitis B virus and alanine aminotransferase, and construct a timeline-aligned serum index sequence matrix. The evolution path construction module analyzes the concentration change direction of hepatitis B surface antigen and hepatitis B virus in the serum index sequence matrix, marks the concentration trend path, records the response direction of alanine aminotransferase at the trend path node, and generates trajectory path structure data. The feature segment extraction module filters trajectory segments with a combination of increasing, stable and turning features in the trajectory path structure data, performs differential labeling, and generates biological change segment data groups; The intervention response classification module compares the consistency between the biological change segment data group and the indicator trajectory combination and segment change label under the intervention operation, removes the intervention sequences with differential expression, and generates the intervention impact relationship table data structure. The fusion structure summarization module statistically analyzes the number of decreasing fragments of hepatitis B surface antigen, the proportion of stable fragments of hepatitis B virus, and the number of fluctuations of alanine aminotransferase in the data structure of the intervention influence relationship table, and constructs a set of summary analysis results for the entire treatment course.

2. The full-course management system based on the fusion analysis of hepatitis B serological and virological indicators according to claim 1, characterized in that: The serum indicator sequence matrix includes hepatitis B surface antigen concentration sequence, hepatitis B virus concentration sequence, and alanine aminotransferase concentration sequence. The trajectory path structure data includes concentration change direction annotation, path node response information, and multi-indicator joint trend type. The biological change segment data group includes increasing characteristic segments, stable characteristic segments, and turning characteristic segments. The intervention influence relationship table data structure includes intervention behavior type, indicator trajectory combination form, and segment change label matching relationship. The full course inductive analysis result set includes the number of surface antigen decreasing segments, the proportion of stable viral segments, and enzyme fluctuation frequency characteristics.

3. The full-course management system based on the fusion analysis of hepatitis B serological and virological indicators according to claim 1, characterized in that: The intervention sequences that eliminate differential expression refer to combinations of indicator trajectories that, under intervention conditions, show inconsistencies with biological change segments.

4. The full-course management system based on the fusion analysis of hepatitis B serological and virological indicators according to claim 1, characterized in that: The labeled concentration trend path refers to tracking and recording the development trend based on the direction of change in serum indicator concentration.

5. The full-course management system based on the fusion analysis of hepatitis B serological and virological indicators according to claim 1, characterized in that, The serum marker acquisition module includes: The data acquisition submodule acquires raw blood samples from multiple treatment time points in the blood collection device, labels the samples according to patient number and time label, sorts the samples according to time sequence, and generates a time series identifier mapping matrix. The indicator extraction submodule, based on the time series identifier mapping matrix, calls the numbered samples to extract the values ​​of hepatitis B surface antigen, hepatitis B virus, and alanine aminotransferase, and after classification and recording, obtains a multi-indicator synchronous sampling data table. The sequence construction submodule aligns similar indicators according to time nodes based on the multi-indicator synchronous sampling data table, integrates the number and indicator dimension, establishes a two-dimensional mapping relationship between indicators and time, and obtains a serum indicator sequence matrix.

6. The full-course management system based on the fusion analysis of hepatitis B serological and virological indicators according to claim 1, characterized in that, The evolution path construction module includes: The concentration trend determination submodule, based on the serum index sequence matrix, calls the hepatitis B surface antigen concentration value and hepatitis B virus concentration value at consecutive time nodes, calculates the difference between adjacent time nodes, and classifies the increase or decrease according to the positive or negative direction of the difference to obtain the concentration change direction sequence. The trend path labeling submodule assigns path numbers to the markers of consecutive time nodes based on the concentration change direction sequence, merges consecutive nodes with the same change direction into a path segment, and binds the change direction with the path segment marker to obtain multiple trend path label sequences. The response trajectory construction submodule calls the multi-segment trend path label sequence to obtain the alanine aminotransferase values ​​of the corresponding nodes, and marks them according to the direction of change between nodes. The marking results are bidirectionally merged with the trend path sequence to establish a multi-index sequence that corresponds to the joint change in concentration and the response direction, and generate trajectory path structure data.

7. The full-course management system based on the fusion analysis of hepatitis B serological and virological indicators according to claim 1, characterized in that, The feature segment extraction module includes: Based on the trajectory path structure data, the trajectory filtering submodule detects the corresponding concentration change tags in all continuous trajectory segments, filters the set of trajectory segments that simultaneously include three change modes: concentration increase, concentration stability, and concentration transition, and adds a structure index and boundary position to each trajectory segment to obtain a composite change segment sequence. The change recognition submodule calls the composite change segment sequence, extracts the index change combination of time nodes in each trajectory segment, calculates the duration and frequency of each type of change combination in the trajectory segment, and filters and classifies according to the set combination weight interval to obtain the index combination feature label set. The difference labeling submodule compares the feature label sequences between adjacent trajectory segments based on the feature label set of the index combination, calculates the label offset rate and sets the feature offset threshold, marks the segment group with the difference degree greater than the offset threshold, adds an identification number, and generates a biological change segment data group.

8. The full-course management system based on the fusion analysis of hepatitis B serological and virological indicators according to claim 1, characterized in that, The intervention response classification module includes: The intervention matching submodule obtains the combination of indicator trajectories for each intervention sequence under the corresponding time period based on the biological change segment data group, calculates the consistency ratio between the trajectory combination and the segment change label, sorts them according to consistency, numbers and classifies them, and establishes an intervention consistency distribution matrix. The differential screening submodule extracts the intervention sequence numbers whose consistency ratio is lower than the intervention response deviation threshold based on the intervention consistency distribution matrix, removes the sequences from the full intervention set, updates the sequence index table and the identifier list, and obtains the high-difference intervention exclusion set; The corresponding table construction submodule calls the remaining intervention sequences outside the high-difference intervention exclusion set, pairs them one-to-one with the corresponding indicator trajectory combinations, constructs mapping entries between intervention behaviors and indicator change paths, and generates a structured two-dimensional record table to obtain the intervention impact relationship table data structure.

9. The full-course management system based on the fusion analysis of hepatitis B serological and virological indicators according to claim 1, characterized in that, The fusion structure summarization module includes: Based on the data structure of the intervention impact relationship table, the indicator feature statistics submodule extracts the indicator path corresponding to each intervention behavior, and calculates the number of all concentration-decreasing fragments in hepatitis B surface antigen, the ratio of the number of static and stable fragments in hepatitis B virus to the total number of fragments, and the number of fluctuations in alanine aminotransferase, to obtain the indicator change feature parameter set. The combined feature construction submodule calls the indicator change feature parameter set, combines the three types of parameter values ​​corresponding to the intervention behavior as independent feature vector entries, constructs a corresponding table based on the intervention behavior number, marks the numerical range and label number under each feature combination, and obtains the intervention feature combination coding table. The treatment course summary generation submodule extracts the combination feature entries of all intervention behaviors within the entire treatment course based on the intervention feature combination coding table, groups them by coding labels and accumulates the number, and establishes a structured set by combining the grouping statistical results to generate a set of full treatment course summary analysis results.

10. The full-course management system based on the fusion analysis of hepatitis B serological and virological indicators according to claim 1, characterized in that: The intervention response deviation threshold is a numerical limit for determining whether the consistency between the intervention sequence and the biological change segment has deviated.

Images