An AIDS antiviral therapy data monitoring method, device and computer

Abstract

The application relates to an AIDS antiviral treatment data monitoring method, device and computer. The method comprises the following steps: acquiring body state information of an abnormal object, and acquiring treatment role information of the abnormal object; performing federal data sharing on the body state information and the treatment role information to obtain an object shared information set; performing medical detection data calculation on the object shared information set to obtain an object abnormal index calculation result; and performing multi-source data comparative analysis on the object abnormal index calculation result to obtain abnormal trend analysis data of the abnormal object. The method can meet the current complex treatment situation and the increasing patient management demand.

Description

Technical Field

[0001] This application relates to the field of disease treatment monitoring technology, and in particular to a method, device and computer for monitoring HIV antiviral treatment data. Background Technology

[0002] Antiretroviral therapy (IRBT) for HIV is a key technology and core component in controlling HIV transmission and improving the quality of life for infected individuals. Its high treatment coverage and success rate are of great significance for effectively controlling the source of infection and reducing new HIV infections and deaths. IBT is primarily undertaken by public hospitals at all levels. However, due to the large number of participating medical institutions, poor information flow, frequent staff turnover, and varying levels of professional competence, the quality of IBT is difficult to maintain at a consistently high level. With the deepening of HIV / AIDS prevention and control efforts, the need for precise monitoring and assessment of treatment quality is becoming increasingly urgent. Traditional monitoring methods, relying on single systems and with limited functionality, can no longer meet the current complex and ever-changing treatment situation and the growing needs of patient management. Summary of the Invention

[0003] Therefore, it is necessary to provide a method, device, and computer for monitoring HIV antiretroviral therapy data that can meet the current complex and ever-changing treatment situation and the growing needs for patient management, addressing the aforementioned technical problems.

[0004] Firstly, this application provides a method for monitoring HIV antiretroviral therapy data, including: Obtain the physical status information of the abnormal object, and obtain the treatment role information for the abnormal object; Federated data sharing is performed on the physical state information and the treatment role information to obtain an object-shared information set; Medical testing data is calculated on the shared information set of the objects to obtain the calculation results of the abnormal indicators of the objects. Multi-source data comparison and analysis are performed on the calculation results of the abnormal indicators of the object to obtain abnormal trend analysis data for the abnormal object.

[0005] Secondly, this application also provides a device for monitoring HIV antiretroviral therapy data, comprising: The object information acquisition module is used to acquire the physical state information of the abnormal object and the treatment role information for the abnormal object. The object information sharing module is used to perform federated data sharing of the body status information and the treatment role information to obtain an object shared information set. An abnormal indicator calculation module is used to perform medical test data calculations on the object's shared information set to obtain the object's abnormal indicator calculation results. The abnormal trend analysis module is used to perform multi-source data comparison and analysis on the calculation results of the abnormal indicators of the object to obtain abnormal trend analysis data for the abnormal object.

[0006] Thirdly, this application also provides a computer, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement any step of a method for monitoring data on HIV antiretroviral therapy.

[0007] The aforementioned method, device, and computer for monitoring HIV antiretroviral therapy data acquires the physical status information of abnormal subjects and combines it with the treatment role information of the abnormal subjects. A shared information set of subjects is generated using federated data sharing. Based on this, medical test data is used to calculate the abnormal index of the subjects. Furthermore, multi-source data comparison and analysis are carried out on the abnormal index calculation results of the subjects to output abnormal trend analysis data for the abnormal subjects. It enables continuous, real-time, and comparable monitoring of antiviral treatment quality in cross-institutional and cross-system data environments. On the one hand, federated data sharing achieves collaborative computing and information complementarity without exposing raw sensitive health data, reducing sample bias caused by data silos and improving the security and generalization ability of evaluation results. On the other hand, medical testing data calculation quickly transforms newly detected events into quantifiable abnormal indicators, significantly shortening the time delay for abnormal identification and intervention. At the same time, multi-source data comparative analysis places individual indicators in reference systems such as peers, superiors and subordinates, and historical baselines for difference identification and trend characterization, improving the interpretability and early warning accuracy of abnormal evolution. This meets the current complex and ever-changing treatment situation and the increasing needs of patient management, providing more timely, consistent, and traceable decision-making basis for clinical treatment adjustments and public health management. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 This is a diagram illustrating the application environment of an HIV antiretroviral therapy data monitoring method in one embodiment. Figure 2 This is a flowchart illustrating a method for monitoring HIV antiretroviral therapy data in one embodiment; Figure 3 This is a structural block diagram of an HIV antiretroviral therapy data monitoring device in one embodiment; Figure 4This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0010] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0011] This application provides a method for monitoring HIV antiretroviral therapy data, which can be applied to, for example... Figure 1 In the application environment shown, terminal 102 communicates with server 104 via a network, and the communication data undergoes anonymization processing. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104. Terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, etc. Server 104 can be implemented using a standalone server or a server cluster consisting of multiple servers.

[0012] In one exemplary embodiment, such as Figure 2 As shown, a method for monitoring HIV antiretroviral therapy data is provided, which can be applied to... Figure 1 Taking the server in the example, the explanation includes the following steps 202 to 208. Wherein: Step 202: Obtain the physical status information of the abnormal object and the treatment role information for the abnormal object.

[0013] Step 204: Federated data sharing is performed on the body status information and treatment role information to obtain the object shared information set.

[0014] Step 206: Calculate medical test data on the shared information set of the objects to obtain the calculation results of the abnormal indicators of the objects.

[0015] Step 208: Perform multi-source data comparison and analysis on the calculation results of the abnormal indicators of the objects to obtain abnormal trend analysis data for the abnormal objects.

[0016] Among them, abnormal objects are treatment subjects or their corresponding case record entities that meet the preset abnormal judgment rules and are included in the monitoring and analysis process, such as AIDS patients.

[0017] Among them, physical status information is medical status data that characterizes the current and changing status of abnormal objects over time. It includes at least the viral load test results and their timestamps, and can be extended to include related data such as immunology, medication and follow-up.

[0018] Among them, the treatment role information is the role identification and permission information of the treatment participants or institutions associated with the abnormal object, which is used to determine the data access boundaries, the identity of the federation node, and the attribution dimension of the comparative analysis.

[0019] Among them, federated data sharing is a data sharing method that achieves cross-agency collaborative computing by having participating nodes upload processed parameters, statistics or intermediate representations and aggregate them without directly exchanging raw sensitive data.

[0020] Among them, the object shared information set is a shared payload set for anomalous objects, which is aggregated by the federated data sharing process and is used to support subsequent real-time calculation and analysis, and does not contain the original plaintext data.

[0021] Among them, the medical testing data calculation is a process of continuously updating the calculation of virological indicators and scores by performing streaming incremental calculations on relevant data such as viral load, triggered by new virological testing events. The medical testing data includes real-time virological data, immunological data, etc.

[0022] Among them, the object anomaly index calculation result is a structured anomaly index output obtained by calculating the virology and related indicators of the abnormal object according to preset rules, including threshold judgment results, trend scores and their derived indicators.

[0023] Among them, multi-source data comparative analysis is an analytical process that involves statistically comparing the abnormal indicators of an object with reference data (same level, superior and subordinate levels, historical baseline, etc.) from different sources or dimensions after calibrating for comparability in order to identify differences.

[0024] Among them, the abnormal trend analysis data is the result data output by multi-source comparative analysis, used to characterize the evolution of abnormal object indicators over time and changes in their relative positions, and includes at least trend characteristics, difference indicators and relative ranking information.

[0025] Specifically, step 202 collects the physical status information of abnormal subjects from the testing system, electronic medical record / follow-up system, or treatment management system through a standardized interface. The physical status information includes at least the viral load test results and their timestamps, and optionally includes immunological data, medication and follow-up records, etc. Subsequently, the abnormal subjects are uniquely identified and associated, and the data is cleaned (duplicate removal, unit conversion, missing value marking, and initial screening of outliers). The treatment role information associated with the abnormal subject is read from the user and institution management database. The treatment role information is used to determine the identity of the data source node, the scope of access control, and the institutional hierarchy for subsequent comparisons, thereby forming a structured input dataset that can be shared and computed later.

[0026] In one specific implementation, a cross-disciplinary professional workshop was organized, inviting HIV treatment experts, medical statisticians, clinical nurses, and public health experts to participate. The indicator collection process was initiated through brainstorming. During the meeting, experts were encouraged to speak freely and share various factors they had discovered in their practical work and research that might affect the quality of HIV antiretroviral therapy. For example, clinical nurses might propose indicators related to patient compliance during treatment, while public health experts might emphasize indicators related to treatment coverage and impact at the community level. The numerous factors proposed by the experts were initially compiled and categorized, removing duplicates or obviously unreasonable content to form a more comprehensive list of potential indicators.

[0027] Next, the Analytic Hierarchy Process (AHP) is used to further determine the final indicators and weights. First, a hierarchical model is constructed, with the quality of HIV antiretroviral treatment as the target layer. Potential indicators are divided into criteria layers according to different categories, such as virology, immunology, molecular biology, and patient behavior. Each criterion layer contains specific potential indicators as the protocol layer.

[0028] Experts were invited to conduct pairwise comparisons of the relative importance of elements at different levels, constructing a judgment matrix. For example, the relative importance of viral load and CD4+ T cell count in reflecting treatment efficacy was compared. By calculating the eigenvectors and the largest eigenvalue of the judgment matrix, the weight vectors of each indicator relative to the target layer were obtained. During this process, the judgment matrix underwent a consistency check; if it failed, the judgment matrix was readjusted until it passed, ensuring the accuracy of the weights. Based on the weight vectors and considering the actual situation, indicators with a significant impact on treatment quality were selected, ultimately determining 30 core indicators for evaluating the quality of HIV antiretroviral therapy.

[0029] The Analytic Hierarchy Process (AHP) was used to determine the weights of each indicator. First, a hierarchical model was constructed, with the quality of HIV antiretroviral treatment as the target layer, 30 assessment indicators as the protocol layer, and several criterion layers in between. For example, the indicators were categorized into virological, immunological, molecular biological, and patient compliance-related categories as criterion layers. Experts were invited to conduct pairwise comparisons of the relative importance of elements across different layers, constructing a judgment matrix. For example, the relative importance of viral load and CD4+ T cell count in reflecting treatment efficacy was compared.

[0030] By calculating the eigenvectors and the largest eigenvalue of the judgment matrix, the weight vectors of each indicator relative to the target layer are obtained. To ensure the accuracy and reliability of the weights, a consistency check is performed on the judgment matrix. If the consistency check fails, the judgment matrix is ​​readjusted until it passes the check. The weights obtained in this way can scientifically and reasonably reflect the relative importance of each indicator in assessing treatment quality.

[0031] Establish a unified data interface to automatically collect patients' basic information (such as age, gender, route of infection, etc.), treatment information (type of medication, medication time, treatment plan adjustment records, etc.), and testing information (viral load, CD4+ T cell count, various biochemical indicators, etc.).

[0032] The collected data undergoes initial screening to remove obviously erroneous or illogical data. For example, data showing a negative viral load or CD4+ T cell count exceeding the normal physiological range is marked and further verified.

[0033] The data is encrypted using the national standard SM4 encryption algorithm. The encrypted information is stored and transmitted in ciphertext form, and only authorized users can recover the true information after obtaining the decryption key, effectively protecting patient privacy.

[0034] To remove noise from the data, filtering algorithms are used. For example, for abnormal fluctuations caused by detection errors, methods such as moving average filtering are used to smooth the data. For missing values, methods such as mean imputation and regression imputation are used to fill in the gaps based on the characteristics and correlations of the data. For data normalization, the data is uniformly mapped to standard intervals such as [0,1] or [-1,1] according to the numerical range and properties of different indicators, to facilitate subsequent analysis and calculation.

[0035] Define different user roles, including provincial business managers, municipal business managers, primary healthcare workers, and data administrators. Assign a unique identifier and set of permissions to each role.

[0036] Provincial-level business management personnel have the highest authority, enabling them to view aggregated data from all medical institutions across the province, including overall treatment quality assessment results and statistical analyses of various indicators. However, they cannot directly modify data for specific patients. They can set some global parameters of the system, such as adjusting the weight range of certain indicators.

[0037] Municipal-level business management personnel have slightly lower authority; they can view data from various medical institutions within the city and conduct comparative analyses within the city, but they also cannot arbitrarily modify detailed patient information. Their main responsibilities include supervising and managing treatment efforts within the city, reporting the city's situation to higher authorities, and relaying instructions from higher authorities to lower levels.

[0038] Primary healthcare workers can only access and manipulate the data of the patients they are responsible for. They can enter the latest treatment information and test results of patients, and view patients' historical treatment records and evaluation reports in order to develop personalized treatment plans for patients.

[0039] Data administrators are responsible for system data maintenance, including data backup, recovery, and data structure adjustments. However, their access to specific patient treatment data is strictly limited; access is only permitted during necessary system maintenance scenarios and after a rigorous approval process.

[0040] By setting up a permission management module in the system, when a user logs in, the system automatically loads the corresponding permission set based on the user's role identifier, restricting the user's access to and operation of data.

[0041] When a user logs in and binds their device, they must provide their mobile phone number, name, ID number, and other information when applying for a system account for the first time. The application form must be stamped with the hospital's official seal. The provincial business administrator will then activate the account after reviewing the approved application form. The account will be bound to the user's mobile phone identification information upon first activation. The system only allows login via mobile APP QR code scanning to ensure data security.

[0042] Step 204: Within each participating institution, the system identifies federated participating nodes and their permission boundaries using treatment role information. Based on local compliance policies, it performs local feature processing or local model / statistic calculations on the body status information, generating a shared payload (e.g., standardized feature vectors, gradient / model parameters, encrypted statistics, or intermediate representations) that does not contain original plaintext data. This shared payload is then encrypted during transmission and authenticated by the nodes before being uploaded to the aggregation end. The aggregation end performs parameter aggregation or consistency fusion based on node weights or sample size, outputting an object-shared information set for anomalous objects. This object-shared information set is used to achieve cross-institutional information complementarity and collaborative computation without directly exchanging original sensitive data.

[0043] In one specific embodiment, in a cross-institutional data sharing scenario, each medical institution, as a participant in federated learning, retains its local data and does not directly share the original data.

[0044] Each participant performs data preprocessing locally, such as feature extraction and data encryption. For example, patient test result data is converted into feature vectors and encrypted using techniques such as homomorphic encryption.

[0045] The parties exchange encrypted model parameters or intermediate results through a secure communication protocol. For example, when training a machine learning model to evaluate treatment quality, each party trains the model locally using its own data, and then sends the trained model gradients and other parameters to other participants.

[0046] Each party updates its local model based on the model parameters received from other participants. Through multiple iterations, each party eventually obtains a global model trained on the data of all participants. However, throughout the entire process, each party cannot access the original data of other participants, achieving "data usable but not visible," thus ensuring data security and privacy.

[0047] Step 206 inputs incremental viral load data from the shared information set of objects as streaming events into the real-time computing engine. The online status is maintained using the abnormal object identifier as the key, and incremental updates are performed according to a preset time window. For each new detection event, virological indicator calculation logic is executed, including threshold determination of viral suppression status, online trend estimation and trend scoring of viral load sequences, and the calculated virologically related abnormal indicators are written into the indicator result database. Simultaneously, the computing engine continuously refreshes the latest indicator status of abnormal objects, ensuring that the calculation results of object abnormal indicators are updated in real time with new detection data and can be immediately accessed for subsequent comparative analysis.

[0048] Step 208 retrieves the calculation results of the abnormal indicators of the abnormal objects from the result library and simultaneously extracts multi-source comparison data (including at least one of the following: comparison data of peer institutions, comparison data of superior and subordinate institutions, or baseline data of historical time intervals). First, cross-source comparability processing is performed on the multi-source data (e.g., calibration mapping based on institutional hierarchy, time baseline, or sample structure) to obtain comparable aligned comparison sequences. Then, statistical comparative analysis is performed on the comparable aligned comparison sequences to output abnormal trend analysis data. The abnormal trend analysis data includes at least relative ranking or quantile information, difference indicators from the comparison baseline, and trend characteristics evolving over time, thereby placing the virological abnormality of the abnormal objects in multiple reference systems for interpretation and trend characterization.

[0049] In one specific embodiment, horizontal comparison among institutions at the same level: When a user selects to perform a horizontal comparison among institutions at the same level, the system extracts 30 evaluation indicator data for each institution from the database. For example, it extracts various indicator data from different designated medical institutions within the same city.

[0050] Using data analysis algorithms, the ranking of each institution on each indicator is calculated, and a comparative report is generated. The report not only displays the indicator values ​​for each institution but also visually shows the differences between each institution and the average level through colors, icons, and other methods. For example, indicators above the average level are marked in green with an upward arrow; indicators below the average level are marked in red with a downward arrow.

[0051] At the same time, the system generates intuitive visualization charts, such as bar charts, to compare and display the same indicator values ​​of various institutions, making it easy for users to quickly understand the performance differences of various institutions on different indicators.

[0052] Vertical Comparison Between Higher and Lower Levels: For vertical comparisons between higher and lower levels of organizations, the system extracts relevant data from higher-level organizations and their subordinate organizations at all levels according to administrative divisions or management hierarchies. For example, data comparisons between provincial-level and municipal-level organizations, and between municipal-level and county-level organizations.

[0053] This analysis examines the differences in treatment quality between higher-level and lower-level institutions, identifying instances where certain indicators perform well at higher-level institutions but are deficient at lower-level institutions. Data mining techniques are then used to pinpoint potential causes of these differences, such as factors related to training intensity and resource allocation.

[0054] Generate longitudinal comparison reports, presenting the comparison results in a combination of tables and graphs. For example, a line graph can be used to show the development trend of different levels of institutions on a certain indicator, helping higher-level institutions to develop targeted management strategies and training plans to improve the treatment quality of lower-level institutions.

[0055] Time-based comparison: The system supports comparing treatment quality in a specific region or medical institution at different time granularities, such as year and month. After the user selects a time range, the system extracts 30 evaluation indicators from the historical database for that time interval.

[0056] Calculate the changes of each indicator at different points in time, such as growth rate and decline rate. Use data analysis algorithms to predict the future trends of the indicators. For example, using time series analysis, analyze data on viral load over the past few months to predict the viral load trend for the next month.

[0057] Generate visual charts comparing various indicators over time, such as line graphs, showing the trends of each indicator over time. Additionally, mark key event points on the charts, such as the timing of treatment plan adjustments and the implementation of new policies, to help users analyze the impact of these events on treatment quality.

[0058] In the aforementioned method for monitoring HIV antiretroviral therapy data, information on the physical status of individuals with abnormalities is acquired and combined with their treatment role information. A shared information set is generated using federated data sharing, and medical testing data is used to calculate abnormality indicators. Further multi-source data comparative analysis is conducted on these abnormality indicator calculations to output abnormality trend analysis data for the individuals with abnormalities. This method enables continuous, real-time, and comparable monitoring of antiretroviral therapy quality in a cross-institutional and cross-system data environment. On the one hand, federated data sharing achieves collaborative computation and information complementarity without exposing raw sensitive health data, reducing sample bias caused by data silos and improving the safety and generalization ability of evaluation results. On the other hand, real-time virological computation quickly transforms newly detected events into quantifiable abnormal indicators, significantly shortening the time delay for abnormality identification and intervention. Simultaneously, multi-source data comparative analysis places individual indicators within reference systems such as peers, superiors, subordinates, and historical baselines for difference identification and trend characterization, improving the interpretability and early warning accuracy of abnormal evolution. This meets the current complex and ever-changing treatment situation and the increasing needs of patient management, providing more timely, consistent, and traceable decision-making basis for clinical treatment adjustments and public health management.

[0059] In an exemplary embodiment, medical testing data is used to calculate the abnormality index of the objects based on the shared information set, including steps 302 to 308. Wherein: Step 302: Perform immune reconstitution calculation on the count change data corresponding to helper T lymphocytes in the object shared information set to obtain the immune reconstitution index.

[0060] Step 304: Based on the immune reconstitution index and the viral load data in the object shared information set, perform online analysis on the viral load change trend of abnormal objects to obtain the corrected viral load change trend.

[0061] Step 306: Perform threshold determination analysis on the viral load data to obtain the viral inhibition rate.

[0062] Step 308: Combine the viral suppression rate with the corrected viral load trend to obtain the calculation results of the abnormal indicators of the target.

[0063] Among them, helper T lymphocytes are a subset of T lymphocytes, represented by CD4⁺ T cells, that reflect the body's immune function status.

[0064] Among them, the count change data is the serialized data of cell counts obtained at different detection time points and their increase or decrease over time.

[0065] Among them, immune reconstitution calculation is a process of calculating the magnitude, speed and stability of immune recovery based on the cell count change sequence and forming an immune recovery evaluation quantity.

[0066] Among them, the immune reconstitution index is a comprehensive indicator or score calculated from immune reconstitution and used to quantify the level of immune function recovery in abnormal subjects.

[0067] Among them, viral load data consists of the HIV viral load values ​​of abnormal objects detected at different time points and their corresponding detection timestamp data.

[0068] Among them, the viral load change trend is the direction and speed characteristics of viral load evolution over time, which is usually represented by a trend curve, slope or trend score.

[0069] Among them, the corrected viral load trend is the result of adjusting the trend interpretation or trend parameters by introducing factors such as the immune reconstitution index on the basis of the viral load trend.

[0070] Threshold determination analysis is an analysis process that compares the detected value with a preset threshold or determination rule to output the qualified / unqualified or abnormal / normal status.

[0071] Among them, the virus inhibition rate is the proportion of the number of tests in which the viral load reaches the inhibition judgment standard within a preset time window to the number of valid tests.

[0072] Specifically, the count change data and sampling timestamps of helper T lymphocytes (CD4⁺ T cells) associated with abnormal objects are extracted from the object-shared information set. Basic data normalization is first performed (unit / dimension unification, merging of duplicate records, missing data marking, outlier removal or truncation). Then, the count change data is organized into a time series according to preset time windows or key event intervals, and characteristic quantities reflecting the level of immune recovery (e.g., recovery amplitude, recovery slope, volatility, or staged decline amplitude) are calculated based on this. The characteristic quantities are synthesized into an immune reconstitution index according to preset mapping rules, and the immune reconstitution index, along with the corresponding time interval and confidence markers (e.g., data density / missing rate), is written into real-time status storage.

[0073] Viral load detection values ​​and timestamps of abnormal objects are extracted from the shared information set. Newly detected cases are input as streaming events into the online trend estimation module. A recursive update method is used to model the trend of the viral load sequence and obtain an initial trend (e.g., online estimation output using recursive least squares, Kalman filtering, or exponential smoothing). The slope, inflection point, or residual of the initial trend is dynamically corrected using the immune reconstitution index as a gating factor or correction factor. This allows the sensitivity and threshold of trend interpretation to be adaptively adjusted as the immune reconstitution status improves or deteriorates. Finally, the corrected viral load change trend is output (which can be implemented as a corrected trend sequence or a corrected trend score), and the trend result is incrementally refreshed each time a new event arrives.

[0074] The viral load detection value is thresholded according to the preset virological judgment rules, and each test is mapped to an "inhibited / uninhibited" status. The status is then summarized according to a preset statistical window. During the summary, the "number of valid tests" can be used as the denominator, and invalid or missing test records can be excluded or counted separately. Then, the proportion of the inhibited state within the window is calculated as the viral inhibition rate, and the statistical information corresponding to the inhibition rate (window length, number of valid samples, and time of the most recent test) is output.

[0075] The viral inhibition rate and the trend of corrected viral load change are aligned to the same evaluation period, and the calculation results of the object anomaly index are generated according to the preset fusion rules. The fusion rules can be in the form of weighted synthesis, piecewise function synthesis or gating combination, to ensure that the "current standard" reflected by the inhibition rate and the "future deterioration or improvement direction" reflected by the trend can jointly contribute to the anomaly judgment. The calculation results of the object anomaly index output after fusion include at least a comprehensive anomaly score or a comprehensive anomaly level, and can simultaneously output sub-results (inhibition rate, trend score, immune reconstitution index) and their timestamps.

[0076] In this embodiment, an immune reconstitution index is obtained by calculating the changes in helper T lymphocyte counts in the shared information set of the subjects. This immune reconstitution index is then incorporated into the online analysis of viral load change trends to obtain corrected trend results. Simultaneously, a viral inhibition rate is determined by thresholding the viral load, and this rate is fused with the corrected viral load change trend to generate the calculation results of the subject's abnormality index. This enables earlier identification and more robust interpretation of virological abnormalities, reduces the risk of misjudgment caused by relying solely on a single threshold or trend, and provides more timely and interpretable quantitative evidence for subsequent continuous monitoring, risk stratification, and intervention decisions.

[0077] In an exemplary embodiment, immune reconstitution calculation is performed on the count change data corresponding to helper T lymphocytes in the object-shared information set to obtain the immune reconstitution index, including steps 402 to 406. Wherein: Step 402: Extract event points from the viral load data in the object shared information set to obtain the immune reconstitution calculation interval.

[0078] Step 404: Perform anchor-constrained phase alignment processing on the count change data within the immune reconstitution calculation interval to obtain the event-aligned cell count change sequence.

[0079] Step 406: Perform gated integral mapping on the event-aligned cell count change sequence based on shared parameters in the object shared information set to obtain the immune reconstitution index.

[0080] Event point extraction is a process of identifying key time points (such as reaching the target, rebound, or mutation point) that meet preset rules from time series data such as viral load.

[0081] Among them, the immune reconstitution calculation interval is a time window or phase interval used to calculate immune reconstitution-related indicators, which is determined by the event point as the boundary.

[0082] Among them, the anchor point constraint phase alignment process is a process of mapping different sampling time axes to a unified event phase axis and aligning the sequence by using event points as anchor points.

[0083] Among them, the event-aligned cell count change sequence is the sequence data of cell count changes with phase that can be directly compared on the same event phase axis after phase alignment by anchor point constraint.

[0084] Among them, shared parameters are model parameters, statistics, or intermediate representations formed through federated data sharing and aggregation, which can be used across nodes and do not contain the original data.

[0085] Among them, gated integral mapping is a process that uses gated weights generated by shared parameters to weight the event-aligned sequence and maps it to a single exponential output through integration / cumulative method.

[0086] Specifically, the viral load detection values ​​and timestamps of abnormal objects are read from the shared object information set. The viral load detection values ​​are then sorted by the timestamp and key event points are identified according to preset event rules. These event rules may include at least one of the following: viral suppression target achievement event points, viral rebound event points, continuous abnormal event points, or rate of change mutation event points. After identifying the event points, an immune reconstitution calculation interval is constructed using the event points as boundaries. For example, a preset time period before and after the event point can be selected to form an analysis window, or the interval between adjacent event points can be selected as the calculation interval. The immune reconstitution calculation interval is then defined by (start and end time, event type, and event confidence marker).

[0087] Within the immune reconstitution calculation interval, helper T lymphocyte count changes and sampling timestamps are extracted. Using the event points obtained in the previous step as anchors, a phase coordinate system is established, mapping the original "time axis" to an "event phase axis." This ensures that cell count changes at different sampling frequencies and detection intervals are aligned around the same event anchor. Cell counts are then resampled or interpolated on the phase axis to ensure that the aligned sequences have consistent alignment positions and comparable sampling granularity near the anchor points. The final output is an event-aligned cell count change sequence, which semantically represents the "immune response trajectory around the same virological event."

[0088] Shared parameters are read from the shared information set of objects, and a gating function or gating weight sequence is generated accordingly. The gating function is used to assign different contributions to different segments of the immune sequence on the event phase axis, thereby highlighting the more critical stages for immune reconstitution and suppressing the influence of noisy segments. The event-aligned cell count change sequence is integrally mapped under the gating weight constraint, compressing the "phase-aligned immune response trajectory" into a single quantifiable output. Specifically, this can be implemented by summing or integrating the gated weighted sequence to obtain an exponential value, and can also include quality markers such as the effective gating interval and effective sample coverage, ultimately yielding the immune reconstitution index.

[0089] In this embodiment, event points are extracted from viral load data in the object shared information set to determine the immune reconstitution calculation interval. Within this interval, cell count change data is anchor-constrained and phase-aligned to form an event-aligned cell count change sequence. Furthermore, based on the shared parameters in the object shared information set, the event-aligned sequence is gated integrally mapped to obtain the immune reconstitution index. This approach can unify the comparison scale of immunological sequences by using key virological events as a reference, reducing biases caused by inconsistent detection frequencies, uneven time intervals, and noise fluctuations. At the same time, the contribution of different phase segments is adaptively adjusted using shared parameters, improving the stability, comparability, and interpretability of the immune reconstitution quantification results.

[0090] In an exemplary embodiment, an immune reconstitution index is obtained by performing a gated integral mapping based on shared parameters in an object-shared information set on the event-aligned cell count change sequence, including steps 502 to 508. Wherein: Step 502: Perform gating parameter transformation on the shared parameters in the object shared information set to obtain the gating function parameters.

[0091] Step 504: Identify stable segments in the event-aligned cell count change sequence to obtain stable segment boundaries.

[0092] Step 506: Based on the gating function parameters and the stable segment boundary, segment gating weights are applied to the event-aligned cell count change sequence to obtain the gating weight sequence.

[0093] Step 508: Perform a weighted integral on the gating weight sequence and the event-aligned cell count change sequence to obtain the immune reconstitution index.

[0094] Among them, gating parameter conversion is the process of converting the shared parameters in the object shared information set into a set of parameters that can directly drive the gating function according to a preset mapping rule.

[0095] Among them, the gating function parameters are a set of parameters used to determine the shape and intensity of the gating function, including threshold, slope, piecewise coefficient or attenuation coefficient, etc.

[0096] Among them, stable segment identification is a process of detecting relatively stable segments based on stability characteristics such as the volatility or rate of change of cell count change sequences.

[0097] Among them, the stable segment boundary is the starting and ending position identifier of the stable segment in the sequence, which is used to limit the range of the interval corresponding to each stable segment.

[0098] Among them, segmented gating weighting is a process of segmenting the sequence based on the boundary of the stable segment and generating and assigning gating weights to each segment according to the gating function parameters.

[0099] Among them, the gating weight sequence is a weight sequence that corresponds point-by-point to the event-aligned cell count change sequence, used to represent the contribution of each phase point to the calculation of the immune reconstitution index.

[0100] The weighted integral is a computational process that involves weighting the gating weight sequence and the cell count change sequence point by point, then summing or integrating them within the calculation interval to obtain the exponential output.

[0101] Specifically, shared parameters related to immune reconstitution (such as model parameters, statistics, or intermediate representations obtained from federated sharing / aggregation) are read from the object shared information set, and their format and version are verified to ensure that the parameters are consistent with the gating function template used in the current immune reconstitution calculation. The shared parameters are converted into gating function parameters according to a preset "parameter mapping table" or "conversion rules." Conversion rules may include dimension alignment (mapping vector / matrix parameters to the scalar or piecewise coefficients required by the gating function), scaling transformation (unifying parameters from different sources to the same numerical domain, such as the logarithmic or standardized domain), constrained projection (projecting parameters to an allowable range to avoid gating instability, such as limiting the gating slope to a preset interval), and parameter combination (combining multiple shared parameters according to rules into gating thresholds, gating gains, attenuation coefficients, or piecewise weighting benchmarks, etc.). After conversion, the gating function parameters are obtained and bound to the abnormal object identifier, calculation interval identifier, and parameter version number for storage. The same set of gating function parameters can be reused traceably during piecewise weighting.

[0102] The sequence of cell count changes aligned to the event (typically including phase point / sampling point indices, aligned cell count values, and optional sampling density markers) undergoes basic quality checks (e.g., removing outliers significantly exceeding physiologically reasonable ranges, marking consecutive missing segments, and verifying the monotonicity of the phase index). Then, local statistics for stability assessment are calculated on the sequence, such as local rate of change, local variance / volatility, and the absolute or relative value of the difference between adjacent points. Based on this, stable segments are identified according to preset stability criteria. These criteria can employ threshold-based rules (a segment is considered stable when local volatility is below a threshold and consistently meets the minimum length requirement) or state transition strategies (dividing the sequence into "stable state / transitional state / volatile state" and determining segments with minimal switching cost). This determines the start and end phase points of each stable interval, outputting a set of stable segment boundaries. These boundaries are used to divide the immune sequence into intervals with different stability characteristics.

[0103] The cell count change sequence is segmented and indexed using stable segment boundaries, dividing the sequence into multiple segments. A gating function parameter is applied to each segment to generate a gating weight generation strategy. Specifically, the gating weight pattern for each segment is determined based on the gating function parameters (e.g., higher and smoother weights are used for stable segments to emphasize reliable intervals, lower or decreasing weights for fluctuating segments to suppress noise, and progressive weights that change with phase to reflect the immune change process for transition segments). A corresponding gating weight value is calculated for each phase point / sampling point within a segment, thus forming a gating weight sequence that corresponds one-to-one with the cell count change sequence. To avoid distortion caused by abnormal weights, consistency constraints are applied to the generated gating weight sequence (e.g., non-negative weights, maximum value limits, intra-segment smoothness constraints, or total weight normalization), but the output remains a single "gating weight sequence." The final gating weight sequence encodes both the adaptive gating strategy resulting from shared parameters and explicitly reflects the constraints of the stable segment structure on weight allocation.

[0104] The gating weight sequence is aligned with the event-aligned cell count change sequence according to the same phase point / sampling point, and then weighted point by point. The cell count value of each point is multiplied by the corresponding gating weight to obtain the weighted contribution value. This value is then accumulated and summed or numerically integrated within the immune reconstitution calculation interval according to a preset integration rule to generate an immune reconstitution index. The integration rule can be based on the phase step size or sampling interval (e.g., if the phase step sizes are unequal, a step size factor is introduced during integration) to ensure that the index has a consistent dimensional interpretation across different sampling densities. After integration, a single immune reconstitution index is output, and quality metadata strongly correlated with this index (e.g., effective phase coverage, stable segment percentage, weight distribution summary, integration interval identifier) ​​can be recorded simultaneously.

[0105] In this embodiment, by mapping the shared parameters in the shared information set of objects to gating function parameters, and identifying the stable segment boundaries on the event-aligned cell count change sequence to guide segmented weighting, a gating weight corresponding to each point of the sequence is generated and weighted integration is performed to obtain the immune reconstitution index. This achieves a combination of "shared parameter-driven adaptive gating" and "stable segment-constrained noise suppression", enabling the immune reconstitution index to maintain a consistent quantification caliber and higher robustness under different sampling frequencies and fluctuation levels. At the same time, the main contribution segments of the index can be clearly traced by the stable segments and weight distribution, thereby improving the credibility and interpretability of the immune reconstitution assessment results.

[0106] In an exemplary embodiment, based on the immune reconstitution index and viral load data in the object-shared information set, the viral load change trend of abnormal objects is analyzed online to obtain a corrected viral load change trend, including steps 602 to 608. Wherein: Step 602: Construct the immune reconstitution index and perform gating coefficients to obtain the immune gating coefficients.

[0107] Step 604: Perform time interval attenuation recalibration on the viral load data to obtain the attenuated and recalibrated viral load sequence.

[0108] Step 606: Perform recursive fitting online analysis on the attenuated recalibrated viral load sequence to obtain the initial viral load change trend.

[0109] Step 608: Based on the immune gating coefficient, perform gating residual correction on the initial viral load change trend to obtain the corrected viral load change trend.

[0110] Among them, the immune gating coefficient is a gating weight factor generated by mapping the immune reconstitution index and used to adjust the sensitivity and correction strength of viral load trend estimation.

[0111] Among them, the time interval decay recalibration is a process of introducing decay weights based on the time interval of adjacent viral load detections and scaling the observation contribution accordingly.

[0112] Among them, the attenuated recalibrated viral load sequence is a time series data sequence for trend estimation formed by recalibrating the viral load observations and their influence weights according to the time interval attenuation rule.

[0113] Among them, recursive fitting online analysis is an online fitting analysis process that updates the trend model parameters incrementally when new observations arrive without requiring a full recalculation.

[0114] The initial viral load trend is the viral load trend result (such as slope, curve or score) obtained by recursive fitting online analysis before the introduction of immune gating residual correction.

[0115] Among them, gated residual correction is a correction process that uses immune gate coefficients to gate and adjust trend residuals or biases and then back-injects to correct the trend results.

[0116] Specifically, the immune reconstitution index and its corresponding timestamp or interval identifier undergo scaling and stabilization processing to ensure they fall within a preset gating range. Based on preset gating mapping rules, the immune reconstitution index is converted into immune gating coefficients. These gating mapping rules can be piecewise or continuous function mappings, used to reduce sensitivity to short-term virological fluctuations when the immune status is good, and to increase the penalty for abnormal changes when the immune status is poor. During the generation process, boundary constraints and smoothing constraints can be applied to the immune gating coefficients to avoid over-amplifying single anomalies. The generated immune gating coefficients are then bound and stored with the anomaly object identifier, the current analysis window, and the parameter version.

[0117] Virus load detection values ​​and timestamps of abnormal objects are extracted from the shared object information set. After sorting by time, the time interval between adjacent detections is calculated, and the time interval is mapped to a decay weight according to a preset decay function. This reduces the contribution of observations with longer intervals to the current trend estimation according to the decay rule, while relatively increasing the contribution of observations with shorter intervals. Each viral load observation is recalibrated, specifically by applying decay weights to the observation value or scaling its influence factor, thus forming a decayed and recalibrated viral load sequence. This sequence also carries the effective contribution information after decay processing, which can reduce the trend estimation bias caused by irregular sampling and improve the stability of online updates.

[0118] The attenuated and recalibrated viral load sequence is used as online input. The state variables for trend estimation are maintained using the abnormal object identifier as the key. A recursive fitting update is performed once each new detection event arrives. The recursive fitting can use recursive least squares, Kalman filtering, or equivalent incremental estimation methods to enable the trend parameters (such as level, slope, or rate of change) to be updated in real time without reviewing all historical data. The output initial viral load change trend can be represented as the trend slope, trend curve prediction, or trend score at the current moment. This initial trend and the corresponding state variables are written into the state storage.

[0119] After obtaining the initial viral load trend, a residual measure is constructed to characterize the deviation between the observed sequence and the initial trend prediction, or the anomalous jumps in trend parameters within a short period. This residual measure is then gated using an immune gating coefficient, which controls the retention, suppression, or penalty intensity of the residual in a multiplicative or additive manner, thus adaptively adjusting the trend correction magnitude under different immune states. Finally, the gated residual is reinjected into the initial trend parameters or trend curve to generate the corrected viral load trend, and the corrected trend result, along with its corresponding gating coefficient summary and residual correction label, is output.

[0120] In this embodiment, immune gating coefficients are generated from the immune reconstitution index, and viral load data is recalibrated by time interval decay. Then, based on the decay-recalibrated viral load sequence, recursive fitting online analysis is performed to obtain the initial viral load change trend. The immune gating coefficients are used to perform gated residual correction on the initial trend to obtain the corrected viral load change trend. This can improve the stability and real-time performance of trend estimation in real-world scenarios with irregular detection time intervals. At the same time, the immune recovery status is introduced into trend interpretation and correction intensity control, reducing the risk of misjudgment caused by short-term noise fluctuations or sparse sampling, and improving the sensitivity and interpretability of virological anomalies.

[0121] In an exemplary embodiment, the initial viral load change trend is corrected using a gating residual based on an immune gating coefficient to obtain a corrected viral load change trend, including steps 702 to 706. Wherein: Step 702: Based on the immune gating coefficient, perform trend gating projection on the initial viral load change trend to obtain the gating trend baseline.

[0122] Step 704: Perform differential analysis on the viral load data and the gating trend baseline to obtain the gating residual sequence.

[0123] Step 706: Perform phase-preserving smoothing on the gated residual sequence to obtain robust gated residuals.

[0124] Step 708: The initial viral load change trend and the robust gating residual are superimposed and corrected to obtain the corrected viral load change trend.

[0125] Trend-gated projection is a process that uses immune gating coefficients to perform a gating transformation on the initial viral load trend to generate a reference trend for residual calculation.

[0126] Among them, the gating trend baseline is a baseline viral load trend sequence or trend model output that has been obtained by trend gating projection and has reflected the influence of immune gating.

[0127] Among them, differential analysis is an analytical process that calculates the difference between the observed viral load at the same time point and the gating trend baseline value to quantify the degree of deviation.

[0128] Among them, the gated residual sequence is a residual time series obtained by difference analysis, arranged in chronological order, and reflecting the deviation of the observation from the gated trend baseline.

[0129] Among them, phase-preserving smoothing is a smoothing process that reduces residual sequence noise while trying to keep the key peaks and valleys and inflection point time positions from shifting.

[0130] Among them, the robust gated residual is the denoised residual result obtained after phase-preserving smoothing, which is used to provide a more stable and reliable correction amount.

[0131] Among them, superposition correction is a process of injecting robust gated residuals back into the initial trend according to preset rules to correct and output the corrected trend.

[0132] Specifically, upon receiving the immune gating coefficient and the initial viral load change trend (which can be represented as a trend curve, a trend prediction value sequence, or a trend parameter containing level and slope terms), the immune gating coefficient is aligned to the time axis or phase axis of the initial trend. Then, according to a preset gating projection rule, the initial viral load change trend is projected and transformed, so that the initial viral load change trend generates a "more conservative or more sensitive" gating trend baseline under the action of the immune gating coefficient. The gating projection rule can be implemented as scaling the trend slope, suppressing / enhancing the response to trend inflection points, or shrinking / expanding the trend prediction interval, so that the gating trend baseline can reflect the moderating effect of immune status on the interpretation of virological trends.

[0133] The gating trend baseline is time-aligned with the viral load observation at the corresponding time point (when the observation does not completely overlap with the baseline node, the corresponding value of the baseline can be obtained by nearest neighbor or interpolation). For each observation point, the difference between the viral load observation value and the predicted value of the gating trend baseline is calculated. This difference is used to characterize the degree of deviation of the current observation from the gating baseline, and the residual sign can be retained to distinguish between "abnormal rise above the baseline" and "improvement decline below the baseline", forming a residual sequence arranged in chronological order.

[0134] Phase-preserving smoothing is applied to the gated residual sequence to reduce high-frequency noise and occasional spikes while minimizing the introduction of time lag or phase shift, thus preserving the temporal location and direction of anomalies. In practice, smoothing operators or filtering strategies with phase-preserving properties can be used on the residual sequence to ensure consistency between the main inflection points and peak / valley positions before and after smoothing, outputting a robust gated residual as a "denoised and time-stable" residual result.

[0135] After aligning the robust gating residuals with the initial viral load trend on the same time axis, the robust gating residuals are injected back into the initial viral load trend according to a preset overlay rule to correct the trend prediction or trend parameters. This ensures that the corrected trend inherits the overall direction of the initial trend while also reflecting the stable deviation after gating and smoothing. After overlay, the corrected viral load trend is output, and a summary of the correction magnitude (such as the slope change before and after correction, and the cumulative contribution of the residuals) and correction labels can be generated simultaneously.

[0136] In this embodiment, by introducing immune-gated projection to form a unified reference baseline during the trend correction process, and explicitly characterizing the deviation between the observation and the baseline using differential residuals, and then obtaining robust residuals through phase-preserving smoothing and superimposing backinjection to complete the correction, the trend estimation error can be separated from the trend ontology and managed in the form of traceable residuals. This ensures that each trend correction has a clear source explanation and an auditable correction magnitude record, thereby facilitating consistency verification and comparability review of trend results under different institutions, different time windows, and different population conditions. At the same time, it provides structured and reusable intermediate results support for threshold determination, adaptive adjustment of alarm thresholds, and subsequent multi-source comparative analysis.

[0137] In an exemplary embodiment, multi-source data comparison and analysis are performed on the calculation results of the object anomaly index to obtain anomaly trend analysis data for the anomaly object, including steps 802 to 806. Wherein: Step 802: Construct cross-source comparability calibration matrices corresponding to each multi-source data in the object anomaly index calculation results.

[0138] Step 804: Based on the cross-source comparability calibration matrix, perform cross-source mapping on each multi-source data to obtain a comparable aligned comparison sequence.

[0139] Step 806: Perform statistical comparative analysis on the comparable aligned comparison sequences to obtain abnormal trend analysis data.

[0140] Among them, multi-source data is a set of indicator data used for comparison and reference from multiple sources such as different organizational levels, different business systems, or different time baselines.

[0141] Among them, the cross-source comparability calibration matrix is ​​a set of calibration parameters or transformation matrix that characterizes the systematic differences between different data sources and is used to transform the data from each source to a uniform comparability scale.

[0142] Cross-source mapping is a process that performs scaling transformation and caliber alignment on the index values ​​or sequences of each data source based on the cross-source comparability calibration matrix to achieve comparability.

[0143] Among them, the comparable aligned comparison sequence is a multi-source indicator time series data that can be directly compared after being mapped across sources and aligned in terms of indicator dimension and time granularity.

[0144] Among them, statistical comparative analysis is an analytical process that calculates statistical results such as ranking, difference measurement and time trend on comparable aligned comparison sequences to output abnormal trend analysis data.

[0145] Specifically, the multi-source data and their corresponding "source identifiers" in the calculation results of the object anomaly index are identified. The multi-source data can come from different organizational levels (same level, superior level, subordinate level), different business systems, or different historical time intervals. Using the index caliber of the object anomaly index calculation results as a unified reference, index dimension alignment and caliber consistency checks are performed on each multi-source data (field mapping, unit conversion, time window consistency). Based on a preset comparability calibration strategy, a cross-source comparability calibration matrix is ​​constructed. The cross-source comparability calibration matrix is ​​used to characterize the systematic difference correction relationship between sources. For example, standardized coefficients are obtained by estimating the mean / variance or quantile baseline between sources, or linear / piecewise mapping coefficients are obtained by fitting historical overlapping samples / common baseline periods between sources. The final output cross-source comparability calibration matrix is ​​usually stored in the form of a set of "source-to-source" mapping coefficients or a transformation matrix from "source to common scale".

[0146] Using the cross-source comparability calibration matrix as the transformation rule, cross-source mapping is performed on the indicator data from each source, transforming the original multi-source indicator vectors or indicator time series to a unified common scale or a unified reference source scale. During the mapping process, the multi-source data is aligned according to the outlier identifier and time axis, and indicator values ​​within the same evaluation period are grouped into comparison samples. Missing or low-quality points are handled using consistent rules (such as missing data marking, imputation, or removal) to ensure that the data from each source can be directly compared at the same time granularity and the same indicator dimension after mapping. After mapping, a comparable aligned comparison sequence is obtained, which is stored in a structured format of "time—source—indicator value / score".

[0147] After obtaining comparable aligned comparison sequences, statistical comparative analysis is performed for each evaluation period or rolling window. The analysis includes at least relative position calculation and difference measurement calculation, such as calculating the ranking / quantile of the abnormal object in the same source set, the difference and difference indicator with the comparison baseline (mean, median or historical baseline), and the trend characteristics across time (slope, volatility, inflection point or duration of continuous deviation). At the same time, the difference results from different sources can be summarized to generate a "trend profile", such as forming a combination of "source difference vector + time trend vector", and outputting abnormal trend analysis data. The abnormal trend analysis data is presented in structured fields (ranking information, difference indicator information, time trend information and its corresponding timestamp / window identifier).

[0148] In this embodiment, by first constructing a cross-source comparability calibration matrix and completing the cross-source mapping alignment of multi-source data accordingly, and then performing statistical comparative analysis on the comparable aligned comparison sequences, the multi-source differences can be shifted from "noise in the analysis stage" to "explicit parameters in the calibration stage". This transforms the comparative analysis process from relying on empirical interpretation to relying on reusable calibration rules. As a result, when data sources are added, organizational levels are adjusted, or indicator calibers are iterated, the comparative capability can be quickly migrated by simply updating the calibration matrix. This significantly reduces the maintenance cost and repetitive development cost of comparative analysis, and improves the scalability and consistency of abnormal trend analysis in different deployment environments.

[0149] In an exemplary embodiment, statistical comparative analysis is performed on comparable aligned comparison sequences to obtain abnormal trend analysis data, including steps 902 to 908. Wherein: Step 902: Perform segmentation processing on the comparable alignment comparison sequence based on key event points to obtain the event segmentation comparison sequence.

[0150] Step 904: Perform segmented change rate difference vectorization on the segmented comparison sequence of events to obtain the event difference vector.

[0151] Step 906: Encode the event difference vector using difference fingerprints to obtain the difference fingerprint.

[0152] Step 908: Based on the difference fingerprint, perform statistical comparison analysis on the event segmentation comparison sequence to obtain abnormal trend analysis data.

[0153] Among them, key event points are time points or phase points in which the state changes significantly during the monitoring process, and serve as event identifiers for the boundaries of the comparison sequence segments.

[0154] Among them, the event segmented comparison sequence is a comparison sequence data organized by event segments, which is formed by dividing the comparable aligned comparison sequence into multiple continuous segments with key event points as boundaries.

[0155] Among them, segmented rate of change difference vectorization is a process of calculating the rate of change characteristics within each event segment and obtaining the differences between sources, and then concatenating them into a vector representation in the segmented order.

[0156] Among them, the event difference vector is a vectorized difference feature composed of the difference in the rate of change of each event segment in a predetermined order, used to characterize the difference pattern before and after the event.

[0157] Among them, differential fingerprinting is an encoding process that maps event difference vectors into a more robust and searchable fingerprint representation through methods such as quantization, bucketing, or hashing.

[0158] Among them, the differential fingerprint is a discretized or fixed-length identifier generated by differential fingerprint encoding, used to represent the differential patterns and trend categories of abnormal objects relative to multi-source references.

[0159] Specifically, a set of key event points for segmentation is determined. These key event points can originate from target achievement / rebound event points in virological testing, mutation points in abnormal indicator calculations, treatment plan change points, or preset time nodes. These key event points are projected onto the time axis or phase axis of comparable aligned comparison sequences. Using these key event points as segmentation boundaries, the sequences are sliced, dividing the entire sequence into multiple continuous segments associated with the events. Each segment is assigned a segment identifier, an event type identifier, and start and end position identifiers, thus forming an event segmented comparison sequence that can be analyzed at an interpretable segment granularity, such as "before the event," "during the event," and "after the event."

[0160] For each segment in the event segmentation comparison sequence, the rate of change characteristics of the abnormal object sequence and the comparison source sequence within that segment are calculated. These rate of change characteristics can be represented by the segment slope, the difference between the first and last segments / duration, the cumulative increment / duration, or the local regression slope, etc. Differences in the rate of change characteristics within the same segment are calculated to obtain the rate of change difference for that segment. These rate of change differences from segment to segment are then concatenated into a vector representation in segment order, thus forming the event difference vector.

[0161] Performing differential fingerprinting on event difference vectors involves mapping continuous numerical vectors into fingerprint representations that offer greater robustness and searchability. During encoding, event difference vectors can be quantized and bucketed to obtain discrete symbol sequences reflecting the direction, intensity, and persistence of differences, or the event difference vectors can be mapped to fixed-length hash / codebook indexes, thus obtaining the differential fingerprint. Differential fingerprints are used to reduce the impact of noise while maintaining the sensitivity of comparison results, enabling statistical comparisons to quickly locate similar difference patterns, identify abnormal patterns, and achieve consistent comparisons across objects and sources, using fingerprints as indexes.

[0162] Using differential fingerprints as the driving input or hierarchical basis for statistical comparative analysis, statistical summarization and comparative calculations are performed on the event segment comparison sequences. For example, weighted aggregation is performed on each segment based on the differential fingerprint matching differential pattern, or quantile estimation is performed on the position of the abnormal object in the comparison group according to the differential fingerprint category. At the same time, statistical comparison results are output at the segment granularity and global granularity, including the ranking / quantile information of the abnormal object in the same level or upper and lower level comparison set, the differential indicator relative to the baseline (higher / lower / closer), and cross-segment trend characteristics (differential accumulation direction, differential amplification / convergence speed, differential change before and after key event points). Finally, abnormal trend analysis data is generated and stored in association with the corresponding event segment identifier and time identifier, so that the display terminal can visualize and trigger alarms according to event segment and time trend.

[0163] In this embodiment, by segmenting comparable aligned comparison sequences according to key event points and vectorizing the differences in the rate of change of each segment, and then encoding the event difference vectors into difference fingerprints and using these fingerprints to drive subsequent statistical comparison analysis, the difference information of high-dimensional and long sequences in multi-source comparisons can be compressed into reusable fingerprint features. This enables the system to not only output single comparison conclusions, but also to quickly retrieve, cluster, and verify difference patterns among different objects, different time windows, and different institutions. This supports the batch identification and reuse of handling strategies for "similar anomaly evolution paths," improving the retrieval efficiency, operational consistency, and experience accumulation capabilities of anomaly trend analysis in large-scale monitoring scenarios.

[0164] In a specific embodiment, 1. Construction and implementation of a multi-dimensional evaluation index system 1.1 Indicator Determination: To monitor the quality of antiretroviral therapy for HIV in a certain region of Yunnan Province, the local health department organized a cross-disciplinary team composed of HIV clinical treatment experts, virologists, nursing experts, public health experts, and medical statisticians. During a brainstorming session, the experts actively raised factors affecting treatment quality based on their experience and expertise. For example, clinical treatment experts pointed out that adverse drug reactions experienced by patients during treatment have a significant impact on treatment adherence and efficacy; nursing experts mentioned that patients' psychological states, such as anxiety or depression, may interfere with the treatment process.

[0165] The numerous factors proposed by experts were compiled, and duplicates and unreasonable parts were removed to form a list of potential indicators. Subsequently, the Analytic Hierarchy Process (AHP) was used to further clarify the final indicators. A hierarchical model was constructed, with the quality of HIV antiretroviral treatment as the target layer, divided into criterion layers according to indicator attributes: virology, immunology, patient behavior, and social support. The virology criterion layer includes potential indicators such as viral load and drug resistance; the immunology criterion layer includes CD4+ T cell count and immune factor levels; the patient behavior criterion layer covers medication adherence and lifestyle regularity; and the social support criterion layer involves family support and the level of community care.

[0166] Experts were invited to conduct pairwise comparisons of the relative importance of elements at different levels, constructing a judgment matrix. For example, the importance of viral load and drug resistance in reflecting treatment efficacy was compared. By calculating the eigenvectors and the largest eigenvalue of the judgment matrix, the weight vectors of each indicator relative to the target layer were obtained, and a consistency test was performed. If the test failed, the judgment matrix was readjusted until it passed, ensuring the accuracy of the weights.

[0167] Based on the weighted vector and combined with the actual situation of HIV / AIDS treatment in the region, 30 core assessment indicators were finally determined. For example, in virology, indicators such as the rate of viral load decline and the frequency of specific drug resistance gene mutations were identified; in immunology, indicators such as changes in the proportion of CD4+ T cell subsets and the levels of immune activation markers were clarified; in patient behavior, indicators such as the rate of timely medication and the degree of active participation in treatment management were included; and in terms of social support, indicators such as family companionship time and participation in community care activities were included.

[0168] 1.2 Weight Allocation: After initially determining the weights using the analytic hierarchy process (AHP), the expert team held another meeting to fine-tune the weights based on the latest research findings and feedback from treatment practices in the region. For example, recent research found that a certain immune factor was closely related to the progression of a patient's disease; after discussion, the experts appropriately increased the weight of the relevant indicators of this immune factor in the evaluation system.

[0169] By combining brainstorming with the analytic hierarchy process, this approach ensures that the evaluation index system can comprehensively reflect the quality of treatment while highlighting key factors, providing a powerful tool for accurately assessing the quality of HIV antiretroviral therapy in the region.

[0170] 2 Data Processing and Security Measures 2.1 Dataset Collection and Cleaning: The region has established a unified data interface, connecting with the information management systems and laboratory testing systems of medical institutions at all levels, including tertiary hospitals, specialized hospitals, and primary healthcare centers. Taking a primary healthcare center as an example, the system automatically collects basic patient information, such as patient Li, 35 years old, male, infected with HIV through sexual transmission, with a history of hypertension; treatment information, such as the current antiviral drug combination of lamivudine, tenofovir, and efavirenz, taken for 2 years, with one dose adjustment due to adverse drug reactions; and testing information, such as the most recent viral load of 50 copies / ml, CD4+ T cell count of 450 cells / μl, and liver function indicators including alanine aminotransferase slightly above the normal range.

[0171] The collected data undergoes initial screening, and abnormal data is marked and verified. For example, if data shows a negative viral load or CD4+ T cell count that is too high or too low, exceeding a reasonable range, the system automatically marks it and notifies relevant personnel for verification.

[0172] Sensitive patient information is encrypted using the national standard SM4 encryption algorithm. Taking patient Li as an example, his name, ID number, home address, and other information are encrypted and stored in the database in ciphertext form. During data transmission, such as when data is transmitted from a primary healthcare center to a higher-level specialist hospital for consultation, it is also transmitted in ciphertext form. Only authorized personnel with the decryption key can restore the true information, effectively protecting patient privacy.

[0173] To address noise in the data, filtering algorithms are used for denoising. For example, for abnormal fluctuations in viral load caused by occasional errors in the testing equipment, a moving average filtering algorithm is used for smoothing. Missing values ​​are imputed based on data characteristics and correlations. For instance, if a patient's CD4+ T cell count is missing in a particular test, the system uses regression imputation based on the patient's previous test data and the average level of similar patients. Simultaneously, data for different indicators are normalized, mapping data of different magnitudes, such as viral load and CD4+ T cell count, to the [0,1] interval for easier subsequent analysis.

[0174] 2.2 Role-Based Access Control (RBAC): The monitoring system in this region clearly defines different user roles and their permissions. For example, Zhang, a municipal-level business manager, has the authority to view summary data on HIV treatment at all medical institutions in the city, including the overall treatment quality assessment scores of each institution and the average values ​​of different indicators, but cannot directly modify specific patient data. He can use the system to analyze the overall trend of treatment quality in the city, providing a basis for formulating citywide treatment policies.

[0175] Wang, a primary care medical worker, can only access and manipulate the data of the patients he is responsible for. For example, if Wang is responsible for the treatment of patient Li, he can enter Li's latest treatment information, such as the most recent symptom feedback and medication adjustment records, and he can also view Li's historical treatment records and evaluation reports in order to develop a personalized treatment plan for Li.

[0176] Data administrator Zhao was primarily responsible for system data maintenance, such as regularly backing up the database and optimizing data storage structure. When operations on patient data were required, such as adjustments to the data format due to system upgrades, Zhao had to go through a strict approval process and obtain authorization from the superior authority before performing any operations. Furthermore, the operation process was strictly recorded and audited by the system.

[0177] This strict role-based access control ensures data security and operational standardization.

[0178] 2.3 Federated learning technology enables data sharing: Several healthcare institutions in the region, such as Hospital A, Hospital B, and Primary Health Care Center C, participated in a data-sharing project based on federated learning. Each institution preprocessed the data locally, for example, converting patient test results into feature vectors and encrypting them using homomorphic encryption.

[0179] Taking the training of a machine learning model to predict patient treatment outcomes as an example, Hospital A, Hospital B, and Primary Healthcare Center C train the model locally using their own data. They then send the resulting model gradients and other parameters to the other participants via a secure communication protocol. Each party updates its local model based on the received parameters. After multiple iterations, each party eventually obtains a global model trained on data from all participants. However, throughout this process, no institution can access the raw data from other institutions. For example, although Hospital A participates in model training, it does not know the specific patient data from Hospital B and Primary Healthcare Center C. It only exchanges encrypted model parameters to collaboratively improve model performance, achieving "data usable but invisible," ensuring data security, and promoting collaboration between different institutions.

[0180] 3. Real-time calculation and comparative analysis functions implemented 3.1 Real-time computing module: The region uses Apache Flink to build a real-time computing platform. When patient Li underwent a new viral load test at a primary healthcare center, with a result of 40 copies / ml, the system immediately sent the data to the streaming computing engine.

[0181] The streaming computing engine pre-defined computational logic for 30 evaluation metrics. Upon receiving new viral load data, it updates virology-related metrics in real time, such as increasing the viral suppression rate from 90% to 92%, and adjusting the viral load trend score accordingly. Simultaneously, it combines this with the most recent change in Li's CD4+ T cell count to comprehensively calculate evaluation metrics related to immune function, such as the immune reconstitution index.

[0182] The calculation results are stored in the database in real time and notified to the front-end display module via a message queue. Wang, a primary care medical worker, logged into the system and immediately saw Li's latest treatment quality assessment on the front-end display module. He found that Li's treatment effect had improved, but a certain indicator related to immune function still needed attention. Therefore, he promptly adjusted the treatment plan, adding an immune-modulating adjuvant therapy to Li's treatment.

[0183] 3.2 Comparative Analysis Module: Horizontal Comparison with Peer Institutions: The region organized a comparative analysis of treatment quality among medical institutions at the same level. Taking three primary healthcare service centers (D, E, and F) as examples, the system extracted data on 30 evaluation indicators. Data analysis algorithms were used to calculate the ranking of each institution on each indicator and generate comparative reports. For example, in the medication adherence indicator, Center D ranked first with an on-time medication rate of 95%; Center E ranked second with an on-time medication rate of 90%; and Center F ranked third with an on-time medication rate of 85%. The report uses color coding: green for Center D, yellow for Center E, and red for Center F, visually demonstrating the differences between each institution and the average level. Simultaneously, the system generated bar charts to clearly show the numerical comparison of the three institutions on various indicators, facilitating understanding of their position among peers and learning from best practices, such as Center D sharing health education methods to improve patient medication adherence.

[0184] Vertical Comparison Between Upper and Lower Levels: A vertical comparison was conducted using municipal hospital G and its three subordinate primary healthcare service centers D, E, and F as an example. System analysis revealed that municipal hospital G significantly outperformed the primary healthcare service centers in controlling virological indicators. Further data mining revealed deficiencies in the accuracy and frequency of testing equipment at the primary healthcare service centers. Based on this, municipal hospital G developed a support plan for the primary healthcare service centers, including providing upgrades to testing equipment and training in testing techniques to improve the quality of treatment at the primary healthcare institutions.

[0185] Time-based comparison: Taking primary healthcare center D as an example, the system supports comparing its treatment quality by month. Users can select to view changes in treatment quality over the past year, and the system extracts data on 30 assessment indicators for the corresponding time interval from the historical database. Time-series analysis algorithms are used to calculate the changes in each indicator. For example, it was found that the average viral load of patients gradually decreased over the past year, while the CD4+ T cell count steadily increased. The system generates graphs showing the trends of these indicators over time and marks key events, such as the introduction of a new antiviral drug combination in May, after which the rate of viral load decline accelerated significantly. Through this time-based comparison, the center can summarize its treatment experience and provide a reference for future treatment plan adjustments.

[0186] Based on the same inventive concept, this application also provides an HIV antiretroviral therapy data monitoring device for implementing the aforementioned method for monitoring HIV antiretroviral therapy data. For example... Figure 3As shown, an HIV antiviral treatment data monitoring device is provided, including: a subject information acquisition module, a subject information sharing module, an abnormal indicator calculation module, and an abnormal trend analysis module. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more HIV antiviral treatment data monitoring device embodiments provided below can be found in the limitations of an HIV antiviral treatment data monitoring method described above, and will not be repeated here.

[0187] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 4 As shown. This computer device includes a processor, memory, input / output interfaces (I / O), and communication interfaces. Those skilled in the art will understand that... Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0188] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0189] In one embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0190] In one embodiment, a computer program product or computer program is provided, the computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and executes the computer instructions, causing the computer device to perform the steps in the above method embodiments.

[0191] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0192] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods.

[0193] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0194] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for monitoring data on HIV antiretroviral therapy, characterized in that, The method includes: Obtain the physical status information of the abnormal object, and obtain the treatment role information for the abnormal object; Federated data sharing is performed on the physical state information and the treatment role information to obtain an object-shared information set; Medical testing data is calculated on the shared information set of the objects to obtain the calculation results of the abnormal indicators of the objects. The step of performing medical testing data calculations on the shared information set of the objects to obtain the calculation results of the abnormal indicators of the objects includes: Immune reconstitution index is obtained by performing immune reconstitution calculation on the count change data corresponding to helper T lymphocytes in the object-shared information set; Based on the immune reconstitution index and the viral load data in the object shared information set, the viral load change trend of the abnormal object is analyzed online to obtain the corrected viral load change trend. The viral load data is subjected to threshold determination analysis to obtain the viral inhibition rate; By combining the virus inhibition rate with the trend of the corrected viral load change, the calculation results of the abnormal index of the object are obtained; The step of performing immune reconstitution calculations on the count change data corresponding to helper T lymphocytes in the object-shared information set to obtain the immune reconstitution index includes: Event points are extracted from the viral load data in the shared information set of the objects to obtain the immune reconstitution calculation interval; Anchor-constrained phase alignment processing is performed on the count change data within the immune reconstitution calculation interval to obtain an event-aligned cell count change sequence; The immune reconstitution index is obtained by performing a gated integral mapping based on shared parameters in the object shared information set on the event-aligned cell count change sequence; The step of performing gated integral mapping on the event-aligned cell count change sequence based on shared parameters in the object shared information set to obtain the immune reconstitution index includes: The shared parameters in the object shared information set are transformed by gating parameters to obtain gating function parameters; The stable segment boundaries are obtained by identifying stable segments in the cell count change sequence aligned with the events. Based on the gating function parameters and the stable segment boundary, the event-aligned cell count change sequence is segmented and gating weighted to obtain a gating weight sequence; The immune reconstitution index is obtained by weighting and integrating the gating weight sequence with the event-aligned cell count change sequence. The step of performing online analysis of the viral load change trend of the abnormal object based on the immune reconstitution index and the viral load data in the object shared information set to obtain the corrected viral load change trend includes: The immune reconstitution index is constructed and a gating coefficient is obtained to obtain the immune gating coefficient; The viral load data is recalibrated at time intervals to obtain the attenuated and recalibrated viral load sequence. The attenuated and recalibrated viral load sequence was recursively fitted and analyzed online to obtain the initial viral load change trend. Based on the immune gating coefficient, the initial viral load change trend is corrected by gating residual correction to obtain the corrected viral load change trend. The step of correcting the initial viral load change trend with a gated residual based on the immune gating coefficient to obtain the corrected viral load change trend includes: Based on the immune gating coefficient, the trend of the initial viral load change is subjected to trend gating projection to obtain the gating trend baseline; Differential analysis was performed on the viral load data and the gating trend baseline to obtain the gating residual sequence; The gated residual sequence is subjected to phase-preserving smoothing to obtain robust gated residuals; The initial viral load change trend and the robust gating residual are superimposed and corrected to obtain the corrected viral load change trend. Multi-source data comparison and analysis are performed on the calculation results of the abnormal indicators of the object to obtain abnormal trend analysis data for the abnormal object.

2. The method according to claim 1, characterized in that, The multi-source data comparison and analysis of the calculation results of the abnormal indicators of the object yields abnormal trend analysis data for the abnormal object, including: Construct a cross-source comparability calibration matrix corresponding to each multi-source data in the calculation results of the object's anomaly index; Based on the cross-source comparability calibration matrix, cross-source mapping is performed on each of the multi-source data to obtain a comparable alignment comparison sequence. Statistical comparative analysis is performed on the comparable aligned comparison sequences to obtain the abnormal trend analysis data.

3. The method according to claim 2, characterized in that, The statistical comparative analysis of the comparable aligned comparison sequences to obtain the abnormal trend analysis data includes: The comparable alignment comparison sequence is segmented based on key event points to obtain an event segmentation comparison sequence; The segmented change rate difference vector is vectorized into the segmented comparison sequence of the events to obtain the event difference vector. The event difference vector is encoded using a difference fingerprint to obtain a difference fingerprint; Based on the difference fingerprint, statistical comparison analysis is performed on the event segmentation comparison sequence to obtain the abnormal trend analysis data.

4. A data monitoring device for HIV antiretroviral therapy, characterized in that, The device includes: The object information acquisition module is used to acquire the physical state information of the abnormal object and the treatment role information for the abnormal object. The object information sharing module is used to perform federated data sharing of the body status information and the treatment role information to obtain an object shared information set. An abnormal indicator calculation module is used to perform medical test data calculations on the object's shared information set to obtain the object's abnormal indicator calculation results. The step of performing medical testing data calculations on the shared information set of the objects to obtain the calculation results of the abnormal indicators of the objects includes: Immune reconstitution index is obtained by performing immune reconstitution calculation on the count change data corresponding to helper T lymphocytes in the object-shared information set; Based on the immune reconstitution index and the viral load data in the object shared information set, the viral load change trend of the abnormal object is analyzed online to obtain the corrected viral load change trend. The viral load data is subjected to threshold determination analysis to obtain the viral inhibition rate; By combining the virus inhibition rate with the trend of the corrected viral load change, the calculation results of the abnormal index of the object are obtained; The step of performing immune reconstitution calculations on the count change data corresponding to helper T lymphocytes in the object-shared information set to obtain the immune reconstitution index includes: Event points are extracted from the viral load data in the shared information set of the objects to obtain the immune reconstitution calculation interval; Anchor-constrained phase alignment processing is performed on the count change data within the immune reconstitution calculation interval to obtain an event-aligned cell count change sequence; The immune reconstitution index is obtained by performing a gated integral mapping based on shared parameters in the object shared information set on the event-aligned cell count change sequence; The step of performing gated integral mapping on the event-aligned cell count change sequence based on shared parameters in the object shared information set to obtain the immune reconstitution index includes: The shared parameters in the object shared information set are transformed by gating parameters to obtain gating function parameters; The stable segment boundaries are obtained by identifying stable segments in the cell count change sequence aligned with the events. Based on the gating function parameters and the stable segment boundary, the event-aligned cell count change sequence is segmented and gating weighted to obtain a gating weight sequence; The immune reconstitution index is obtained by weighting and integrating the gating weight sequence with the event-aligned cell count change sequence. The step of performing online analysis of the viral load change trend of the abnormal object based on the immune reconstitution index and the viral load data in the object shared information set to obtain the corrected viral load change trend includes: The immune reconstitution index is constructed and a gating coefficient is obtained to obtain the immune gating coefficient; The viral load data is recalibrated at time intervals to obtain the attenuated and recalibrated viral load sequence. The attenuated and recalibrated viral load sequence was recursively fitted and analyzed online to obtain the initial viral load change trend. Based on the immune gating coefficient, the initial viral load change trend is corrected by gating residual correction to obtain the corrected viral load change trend. The step of correcting the initial viral load change trend with a gated residual based on the immune gating coefficient to obtain the corrected viral load change trend includes: Based on the immune gating coefficient, the trend of the initial viral load change is subjected to trend gating projection to obtain the gating trend baseline; Differential analysis was performed on the viral load data and the gating trend baseline to obtain the gating residual sequence; The gated residual sequence is subjected to phase-preserving smoothing to obtain robust gated residuals; The initial viral load change trend and the robust gating residual are superimposed and corrected to obtain the corrected viral load change trend. The abnormal trend analysis module is used to perform multi-source data comparison and analysis on the calculation results of the abnormal indicators of the object to obtain abnormal trend analysis data for the abnormal object.

5. A computer, comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 3.

Images