Gestational diabetes mellitus serum marker detection data analysis system

Abstract

The invention relates to the technical field of gestational diabetes mellitus detection, and discloses a gestational diabetes mellitus serum marker detection data analysis system. The system comprises a serum biochemical data acquisition module, a detection device is used for acquiring a serum sample of a pregnant woman, and a serum biochemical data set containing blood sugar concentration, insulin level and glycosylated hemoglobin indexes is constructed; a multi-dimensional index integration module combines a historical detection time sequence and the physiological parameter information of the current detection time period to generate an integrated data set containing time dimension correlation features; the abnormal risk analysis module processes the time dimension correlation features in the integrated data set, and calculates the risk deviation value of gestational diabetes in the current detection period; when the risk deviation value is abnormal, the risk grading module generates a plurality of candidate intervention classification schemes based on the serum biochemical data set, and outputs the intervention influence degree of each candidate scheme; and an intervention scheme generation module screens the candidate scheme with the lowest intervention influence degree and generates an optimal intervention classification scheme in the current detection time period.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of gestational diabetes mellitus detection, in particular to a gestational diabetes mellitus serum marker detection data analysis system. BACKGROUND

[0002] As a common metabolic disease during pregnancy, gestational diabetes mellitus, if not timely discovered and effective intervention measures taken, will not only increase the probability of pregnant women developing complications such as gestational hypertension and polyhydramnios, but also may have adverse effects on the growth and development of the fetus, such as causing fetal macrosomia, fetal distress, and even affecting the long-term health of mother and infant. At present, serum marker detection for gestational diabetes mellitus has become one of the important means of clinical monitoring, but the existing detection data analysis methods still have many limitations.

[0003] In the serum data collection link, most detection methods only focus on a single or a few serum indicators, such as only detecting blood glucose concentration or insulin level, without comprehensively integrating blood glucose concentration, insulin level and glycosylated hemoglobin indicators, resulting in a single dimension of the constructed serum dataset, which is difficult to fully reflect the metabolic state of pregnant women. Such a single-dimensional dataset makes the subsequent analysis process lack sufficient basic data support, and cannot accurately capture the key information in the development process of gestational diabetes mellitus.

[0004] From the perspective of data integration, the existing technology often ignores the influence of time dimension on detection data. Pregnancy is a dynamic process, and the serum biochemical indicators of pregnant women will change with the increase of gestational weeks. The trend of changes in indicators under different historical detection time sequences is of great significance for judging the health status at the current detection period. However, the current data integration method is mostly limited to simple aggregation of detection data at the same time point, without combining historical detection time sequences and physiological parameter information of the current detection period, which cannot generate integrated datasets containing time dimension correlation characteristics, making it difficult to analyze the change rule of serum indicators of pregnant women from a dynamic perspective, and further affecting the accurate judgment of the risk of gestational diabetes mellitus.

[0005] In terms of abnormal risk analysis, the existing system usually only compares the single serum indicator value at the current detection period with the standard value to determine whether there is an abnormal risk, without further processing the time dimension correlation characteristics in the integrated dataset. This analysis method cannot fully consider the change trend of serum indicators at different time stages, and can only reflect the indicator state at the current time point, making the calculated risk deviation value of gestational diabetes mellitus lack comprehensiveness and accuracy, and prone to misjudgment or omission, which makes it difficult to accurately identify potential abnormal risks.

[0006] In the risk intervention scheme generation link, the prior art often generates a fixed intervention scheme based on limited serum data when the risk deviation value is found to be abnormal, without considering the physiological parameter differences of different pregnant women at the current detection period. Due to the individual differences in physiological parameter information of different pregnant women, when the fixed intervention scheme is applied to all pregnant women, adverse effects may occur due to the mismatch of physiological parameters. At the same time, the prior art does not calculate the synergistic interference coefficient when the physiological parameter information at the current detection period is switched to the reference state, and cannot evaluate the intervention influence degree of each candidate intervention classification scheme on the detection analysis process, resulting in that the generated intervention scheme lacks pertinence and rationality, and is difficult to meet the actual needs of different pregnant women, and cannot provide reliable support for effective intervention of gestational diabetes mellitus. SUMMARY

[0007] The purpose of the present application is to provide a gestational diabetes mellitus serum marker detection data analysis system to solve the problems raised in the background art.

[0008] To achieve the above-mentioned purpose, the present application provides a gestational diabetes mellitus serum marker detection data analysis system, which comprises:

[0009] A serum biochemical data acquisition module acquires serum samples of pregnant women through a detection device to construct a serum biochemical data set containing blood glucose concentration, insulin level and glycated hemoglobin index;

[0010] A multi-dimensional index integration module generates an integrated data set containing time dimension correlation characteristics based on the serum biochemical data set, combined with historical detection time series and physiological parameter information at the current detection period;

[0011] An abnormal risk analysis module processes the time dimension correlation characteristics in the integrated data set to calculate the gestational diabetes mellitus risk deviation value based on the serum biochemical data set at the current detection period;

[0012] A risk level division module generates a plurality of candidate intervention classification schemes based on the serum biochemical data set when the gestational diabetes mellitus risk deviation value is abnormal, and calculates the synergistic interference coefficient when the physiological parameter information at the current detection period is switched to the reference state combined with the integrated data set, and outputs the intervention influence degree of each candidate intervention classification scheme on the detection analysis process;

[0013] An intervention scheme generation module selects the candidate intervention classification scheme with the lowest intervention influence degree to generate the best intervention classification scheme at the current detection period.

[0014] Preferably, the serum biochemical data set contains blood glucose concentration, insulin level and glycated hemoglobin values corresponding to consecutive detection time stamps;

[0015] The historical detection time series includes the fluctuation trends of each serum biochemical indicator within the previous three detection cycles;

[0016] The physiological parameters include body mass index, gestational age, and basal metabolic rate for the current testing period.

[0017] Preferably, the time-dimensional correlation features of the integrated dataset include the time-varying correlation strength between blood glucose concentration and insulin level, and the offset of glycated hemoglobin from historical fluctuation trends;

[0018] The method for calculating the risk deviation of gestational diabetes is as follows: extract the body mass index deviation from the physiological parameter information of the current testing period, and combine it with the cumulative deviation of blood glucose concentration from the standard threshold in the serum biochemical dataset for weighted fusion.

[0019] Preferably, when the risk level classification module generates a candidate intervention classification scheme, it identifies the abnormal boundary nodes of each indicator in the serum biochemical dataset, marks the warning intervals corresponding to each abnormal boundary node in the detection and analysis path, and records the safe connection path between the current detection location and each warning interval as the candidate intervention area.

[0020] The calculation basis of the cooperative interference coefficient is the ratio of the total number of operation instructions required to switch physiological parameter information to the baseline state in the historical record to the number of operation instructions in the current switching process.

[0021] Preferably, the method for obtaining the intervention impact includes: calculating the gestational diabetes risk deviation value corresponding to a specific candidate intervention classification scheme, and combining the collaborative interference coefficient and the time dimension correlation feature distribution density of the integrated dataset in the scheme path.

[0022] Preferably, the system further includes:

[0023] The dynamic update module updates the detection and analysis paths in the database in real time according to the optimal intervention classification scheme, and generates detection log information based on the update results;

[0024] When the dynamic update module updates the detection and analysis path, it simultaneously records the mapping relationship between the historical detection time series and the serum biochemical dataset of the current detection period.

[0025] Preferably, the detection log information includes a detection timestamp, the best intervention classification scheme used, the serum biochemical dataset before the update, the integrated dataset, and a snapshot of physiological parameter information.

[0026] Preferably, the system further includes a baseline state storage module, which pre-stores baseline body mass index range, baseline gestational age range, and baseline metabolic rate threshold for physiological parameter information.

[0027] Preferably, the system further includes an intervention program library, which stores the execution parameters of dietary control programs, exercise prescription programs, and drug intervention programs corresponding to different risk levels.

[0028] Preferably, when generating the integrated dataset, the multidimensional index integration module adds phase synchronization markers for time-series blood glucose concentration data and insulin level data.

[0029] Compared with the prior art, the beneficial effects of the present invention are:

[0030] This gestational diabetes mellitus serum biomarker detection data analysis system, in the serum biochemical data collection stage, acquires serum samples from pregnant women through detection equipment and constructs a serum biochemical dataset including blood glucose concentration, insulin level, and glycated hemoglobin indicators. Compared with the traditional collection method that only focuses on a single or a few serum indicators, it can more comprehensively collect key information reflecting the metabolic status of pregnant women, providing richer and more comprehensive basic data for subsequent data analysis. It can more clearly present the overall situation of serum biochemical indicators of pregnant women and provide a more complete information source for subsequent analysis work.

[0031] The multidimensional indicator integration module, based on the constructed serum biochemical dataset, further combines historical test time series and physiological parameter information from the current test period to generate an integrated dataset containing time-related features. This integration method breaks through the limitations of traditional data integration, which is limited to summarizing data at the same point in time. It fully considers the dynamic characteristics of serum indicators in pregnant women over time, correlates historical test data with current test data, and incorporates physiological parameter information. This results in an integrated dataset that not only contains multidimensional serum indicators but also possesses time-related correlations. The integrated dataset generated by this module can more accurately reflect the changing trends of serum biochemical indicators in pregnant women at different gestational weeks, as well as the actual situation of the indicators under the current physiological state. This provides more valuable data for subsequent abnormal risk analysis and helps to more accurately capture the potential patterns in the development of gestational diabetes.

[0032] The anomaly risk analysis module processes the time-related features of the integrated dataset to calculate the risk deviation value of gestational diabetes based on the serum biochemistry dataset for the current testing period. This module departs from the traditional method of comparing indicators at a single time point with standard values, instead delving into the information contained in the time-related features and fully considering the impact of changes in serum indicators at different time stages on the current risk assessment. The risk deviation value calculated in this way can more comprehensively and accurately reflect the risk status of pregnant women with gestational diabetes during the current testing period, effectively reducing misjudgments or omissions caused by focusing only on data from a single time point, and helping medical staff to identify potential anomalies more promptly.

[0033] When gestational diabetes risk deviations are abnormal, the risk grading module generates multiple candidate intervention classification schemes based on the serum biochemistry dataset. Simultaneously, it calculates the co-interference coefficient when the physiological parameters of the current testing period are switched to the baseline state, combining the integrated dataset, and outputs the intervention impact of each candidate classification scheme on the testing and analysis process. This module not only provides multiple intervention options, avoiding the limitations of traditional fixed intervention schemes, but also deeply analyzes the potential impact of different intervention schemes on the testing and analysis process by calculating the co-interference coefficient. Medical staff can clearly understand the applicability and potential impact of different schemes based on the intervention impact of each candidate scheme, providing detailed reference information for subsequent selection of appropriate intervention schemes, making the selection of intervention schemes more scientific and targeted.

[0034] The intervention plan generation module filters out candidate intervention classification plans with the lowest intervention impact, generating the optimal intervention classification plan for the current testing period. Because the screening process prioritizes the lowest intervention impact, it ensures that the generated intervention plan effectively addresses the risks of gestational diabetes while minimizing interference with the testing and analysis process. This reduces the adverse effects of inappropriate intervention plans on the accuracy of subsequent test data and the reliability of analysis results. The optimal intervention classification plan generated in this way is better suited to the specific circumstances of pregnant women at the current testing period, meeting the individual needs of different pregnant women. This provides more rational and effective support for clinical intervention in gestational diabetes, helping medical staff develop more practical intervention strategies for pregnant women and ensuring maternal and infant health. Attached Figure Description

[0035] Figure 1 This is a timing diagram of the gestational diabetes mellitus serum biomarker detection data analysis system described in this invention;

[0036] Figure 2 A flowchart for integrating serum biochemical data with physiological parameters;

[0037] Figure 3 A flowchart for generating candidate intervention schemes and calculating synergistic interference;

[0038] Figure 4 This is a flowchart for analyzing gestational diabetes mellitus detection data, which includes a baseline state storage module. Detailed Implementation

[0039] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0040] Please see Figure 1 This invention provides a data analysis system for detecting serum biomarkers of gestational diabetes mellitus, the overall implementation of which is as follows:

[0041] This system acquires serum samples from pregnant women through a serum biochemistry data acquisition module, detecting blood glucose concentration, insulin levels, and glycated hemoglobin levels to construct a serum biochemistry dataset. A multidimensional indicator integration module, based on this serum biochemistry dataset and combining historical testing time series and physiological parameter information from the current testing period, generates an integrated dataset containing time-related features. An anomaly risk analysis module processes the time-related features in the integrated dataset, calculating the gestational diabetes risk deviation value based on the serum biochemistry dataset for the current testing period. When the gestational diabetes risk deviation value is abnormal, the risk level classification module generates multiple candidate intervention classification schemes based on the serum biochemistry dataset and calculates the co-interference coefficient when the physiological parameter information for the current testing period is switched to the baseline state, outputting the intervention impact of each candidate intervention classification scheme on the testing and analysis process. The intervention scheme generation module selects the candidate intervention classification scheme with the lowest intervention impact and generates the optimal intervention classification scheme for the current testing period. This system automates data analysis and intervention scheme generation, improving the efficiency and accuracy of gestational diabetes risk management.

[0042] Example 1: See Figure 2 This embodiment involves the construction of a serum biochemistry dataset, the integration of historical test time series, the processing of physiological parameter information, and the generation of time-related features. The serum biochemistry data acquisition module obtains serum samples from pregnant women using testing equipment, following standardized clinical operating procedures. The collected samples are tested using a biochemical analyzer to obtain values ​​for blood glucose concentration, insulin level, and glycated hemoglobin. Each test result is associated with consecutive test timestamps, forming time series data. The timestamps record the year, month, day, hour, and minute of the test, ensuring the time accuracy of the data. The serum biochemistry dataset is stored in a structured format, containing fields such as: test ID, timestamp, blood glucose concentration (unit: mmol / L), insulin level (unit: pmol / L), and glycated hemoglobin (unit: %). The data is stored in a relational database, with indexes optimizing query efficiency.

[0043] The processing of historical testing time series data is based on data from the previous three testing cycles. Each testing cycle is defined as a complete serum sample testing event, typically spaced four weeks apart. The calculation of fluctuation trends involves statistically analyzing the mean, standard deviation, and linear regression slope of each serum biochemical indicator within each cycle. For example, the fluctuation trend of blood glucose concentration is characterized by calculating the mean, standard deviation, and slope of all blood glucose concentration values ​​over time within the cycle. The fluctuation trends of insulin levels and glycated hemoglobin are calculated using the same method. Historical data is retrieved from a database, sorted chronologically, and trend features are extracted using time series analysis algorithms (such as moving averages or exponential smoothing). These trend data are stored in separate tables and correlated with the current testing data.

[0044] Physiological parameters include body mass index (BMI), gestational age, and basal metabolic rate (BMR) for the current testing period. BMI is calculated by dividing weight (kg) by the square of height (m). Weight and height data are obtained using a digital scale and height measuring device, with an accuracy requirement of ±0.1 kg for weight and ±0.01 m for height. Gestational age is calculated based on the date of the last menstrual period and the current date, expressed in weeks and rounded to one decimal place. BMR is measured using an indirect caloric analyzer, expressed in kcal / day, under fasting and resting conditions. These parameters are integrated into the dataset, forming a subset of physiological parameters, and are linked to serum biochemical data via test IDs.

[0045] The integrated dataset was generated by combining serum biochemistry datasets, historical detection time series, and physiological parameter information. Time-dimensional correlation features included the time-varying correlation strength between blood glucose concentration and insulin levels, and the offset of glycated hemoglobin from historical fluctuation trends. The time-varying correlation strength was calculated using a sliding window method, with the window size set to three consecutive detection timestamps. Within each window, the correlation coefficient between blood glucose concentration and insulin levels was calculated, adjusting for the effects of time lag. The formula is:

[0046]

[0047] Where S t G represents the time-varying correlation strength. i I represents the blood glucose concentration value at the i-th time point within the window. i This represents the insulin level value at the i-th time point within the window. This represents the average blood glucose concentration within the window. This formula represents the average insulin level within the window, where n represents the number of data points within the window (fixed at 3), λ represents the time decay coefficient (set to 0.1), and Δt represents the time difference (in hours) between the current window's center time point and the reference time point. This formula quantifies the dynamic relationship between blood glucose and insulin over time, taking into account the time decay effect.

[0048] The offset of glycated hemoglobin (HbA1c) from historical fluctuations is calculated by comparing the current HbA1c value with the average of three historical testing periods. The historical average is calculated as the arithmetic mean of the HbA1c values ​​over the three periods. The offset formula is:

[0049]

[0050] Among them O h H represents the offset. c Indicates the current glycated hemoglobin level, H h σ represents the historical average. h This represents the standard deviation of historical data. After standardization, this value reflects the degree of deviation of the current value from the historical pattern.

[0051] The risk deviation for gestational diabetes is calculated based on body mass index (BMI) offset and the cumulative deviation of blood glucose concentration from a standard threshold. The BMI offset is calculated as the difference between the current BMI and the median of the baseline BMI range. The baseline range is set at 18.5-24.9 kg / m². 2 The median value is 21.7 kg / m³. 2 The offset formula is B. o =|B c -21.7|, where B c This represents the current body mass index (BMI). The cumulative deviation of blood glucose concentration from the standard threshold is calculated by summing all blood glucose concentration measurements exceeding the standard threshold (5.1 mmol / L) within the current testing period. The standard threshold is based on clinical guidelines. The formula for the cumulative deviation is: Where m represents the number of measurements in the current detection period, G j This represents the value of the j-th measurement.

[0052] The weighted fusion uses a linear combination, with weighting coefficients set based on clinical importance. The risk deviation formula is R = w1·B. o +w2·D g The weights w1 and w2 are determined by expert consultation and literature review, emphasizing the dominant role of blood glucose deviation. The calculation results are normalized to the 0-1 range for easier comparison. The entire implementation is achieved through software modules, with automated data processing requiring no manual intervention.

[0053] Example 2: See Figure 3This embodiment relates to the process of generating candidate intervention classification schemes by a risk level classification module, and the calculation method of the co-interference coefficient. This module is activated when the risk deviation value of gestational diabetes mellitus is determined to be abnormal, and operates based on a serum biochemistry dataset. The first step in generating candidate intervention classification schemes is to identify the abnormal boundary nodes of each indicator in the serum biochemistry dataset. The identification of abnormal boundary nodes relies on pre-set clinical thresholds and statistical control limits. For blood glucose concentration, an abnormal boundary node is defined as a time point where the concentration value continuously exceeds the standard threshold of 5.1 mmol / L, and it must meet the condition that two consecutive detection time points are outside this range. Abnormal boundary nodes for insulin levels are determined by detection points where the concentration value is below the lower limit of normal (2.5 pmol / L) or above the upper limit (25 pmol / L), also requiring data support from two consecutive time points. Abnormal boundary nodes for glycated hemoglobin are judged by a value exceeding 6.5%, and two of the most recent three detection time points must exceed this threshold. The identification of these nodes is automatically performed by an algorithm that scans the time series data in the serum biochemistry dataset and applies a sliding window technique to detect continuous abnormal patterns.

[0054] After identifying abnormal boundary nodes, the system marks the corresponding warning intervals for each abnormal boundary node in the detection and analysis path. The detection and analysis path is a virtual data structure, representing the trajectory of the detection index over time in the form of a directed graph. Nodes represent the detection values ​​at specific time points, and edges represent possible paths of numerical change. The warning interval is defined as a region extending a certain range before and after the abnormal boundary node along the time axis. For abnormal blood glucose concentration nodes, the warning interval covers a 12-hour period before and after the node; for abnormal insulin levels, the warning interval is 8 hours before and after; and for abnormal glycated hemoglobin, the warning interval extends to 24 hours before and after. These interval ranges are set based on the temporal characteristics of index fluctuations in clinical practice. Within each warning interval, the system sets different levels of risk indicators, divided into three levels from low to high risk, identified by color codes: yellow for mild warning, orange for moderate warning, and red for high warning.

[0055] The safe connection paths between the current detection location and each warning interval are denoted as candidate intervention regions. The current detection location refers to the coordinates of the latest detection time point in the analysis path. The generation of safe connection paths employs the shortest path algorithm from graph theory, starting from the current detection location and avoiding all nodes marked as warning intervals to find safe routes to the boundaries of each warning interval. The generation of these paths considers multiple constraints, including path length, time interval, and numerical gradient changes. Each candidate intervention region corresponds to a specific intervention direction, such as a region targeting blood glucose control, insulin regulation, or glycated hemoglobin management. The system generates an intervention classification scheme for each candidate intervention region, including the target indicator, adjustment magnitude, and time frame.

[0056] The collaborative interference coefficient is calculated based on the ratio of the total number of operational instructions required to switch physiological parameter information to the baseline state in historical records to the number of operational instructions in the current switching process. Historical records are stored in the system database and contain physiological parameter adjustment data for all past test cases. Operational instructions refer to standardized commands executed to adjust body mass index, gestational age, or basal metabolic rate to the baseline state. For example, for body mass index adjustment, operational instructions might include "reduce daily calorie intake by 200 kcal" or "increase moderate-intensity exercise by 30 minutes"; gestational age adjustment instructions involve pregnancy management measures; and basal metabolic rate adjustment instructions include metabolic regulation recommendations. The total number of historical operational instructions is obtained by statistically analyzing the cumulative number of instructions used to successfully achieve the baseline state switch in all similar cases, and the average value is used in the calculation.

[0057] The number of operation commands for the current switching process is obtained through simulation calculation. Based on the difference between the current physiological parameters and the baseline state, the system generates a theoretical sequence of operation commands. For example, if the current body mass index is higher than the median of the baseline range, the system will generate a series of weight loss commands; if the basal metabolic rate is low, it will generate metabolic boosting commands. The number of commands is calculated based on standardized conversion rules, with each command corresponding to a fixed adjustment unit. The proportional relationship is calculated using division, with the historical average number of commands as the denominator and the current simulated number of commands as the numerator. This proportionality coefficient reflects the complexity of the current switching process relative to the historical average level; a value greater than 1 indicates that the current switching requires more operations, while a value less than 1 indicates that fewer operations are required.

[0058] The entire implementation process is executed through a dedicated algorithm engine that integrates graph computation, time series analysis, and rule-based reasoning. Data flow is fully automated, forming a continuous processing pipeline from anomaly node identification to candidate region generation and interference coefficient calculation. All intermediate results and final outputs are stored in the system cache for subsequent modules to access. No manual judgment is relied upon during implementation; all decisions are based on algorithms and predefined rules, ensuring consistency and objectivity. The system regularly updates the historical database, incorporating new successful switchover cases to keep the benchmark for calculating the collaborative interference coefficient up-to-date.

[0059] Example 3: This example relates to a method for obtaining the degree of intervention impact. This method is the core calculation process of the risk level classification module outputting the degree of intervention impact of each candidate intervention classification scheme on the detection and analysis process. The degree of intervention impact, as a quantitative indicator, is used to evaluate the expected impact of each candidate intervention classification scheme on the current state of the system during implementation. Its calculation integrates multiple data sources and feature indicators.

[0060] The assessment of intervention impact begins with a simulation of a specific candidate intervention classification scheme. The system first applies the rules and parameters defined by the scheme to the current serum biochemistry dataset, generating a adjusted simulation dataset. For example, if the candidate scheme is a dietary control scheme for glycemic control, the system uses a pre-established physiological metabolic model to predict the trend of blood glucose concentration changes based on the carbohydrate intake restrictions and meal schedule specified in the scheme. The simulation process considers individual current physiological parameters, such as body mass index and basal metabolic rate, to ensure personalized predictions. After the simulation, the system recalculates the risk deviation value for gestational diabetes based on the adjusted simulation dataset. The recalculation method remains consistent with the initial calculation, extracting the body mass index offset and cumulative blood glucose concentration deviation from the simulation data and weighting them using the same weighting coefficients to obtain a corrected risk deviation value. This corrected value reflects the expected change in risk level after implementing the candidate scheme.

[0061] The system obtains the cooperative interference coefficient corresponding to the candidate solution from the risk level classification module. This coefficient has already been calculated in the preceding process and is directly read from the unresolved cache. The cooperative interference coefficient characterizes the relative effort required to adjust the current physiological parameters to the baseline state, and its value directly affects the complexity and resource consumption of the intervention.

[0062] The system calculates the distribution density of temporal correlation features in the integrated dataset along the candidate intervention classification scheme path. The scheme path refers to the sequence of nodes and edges covered by the candidate scheme in the detection and analysis path. Temporal correlation features include the time-varying correlation strength between blood glucose concentration and insulin level, and the offset of glycated hemoglobin from historical fluctuation trends. The distribution density is calculated by analyzing the distribution of these feature values ​​along the scheme path. The system divides the scheme path into several continuous intervals, and statistically analyzes the frequency and range of feature values ​​within each interval. The distribution density is quantified using a kernel density estimation method, employing a Gaussian kernel function to smooth the distribution of feature values ​​along the path, resulting in a continuous density function. Regions with high density values ​​indicate a high concentration of temporal correlation features along that path segment, potentially suggesting that the region is more sensitive to or unstable to intervention measures.

[0063] The impact of the intervention is calculated using a composite formula that integrates the three elements mentioned above: the adjusted risk deviation, the synergistic interference coefficient, and the distribution density of the time-dimensional correlation features. The formula is designed as follows:

[0064]

[0065] Where I impact R represents the degree of impact of the intervention and is a dimensionless scalar value. adj This represents the risk deviation for gestational diabetes recalculated after implementing the candidate intervention classification scheme. This value is obtained through a weighted fusion of body mass index offset and cumulative blood glucose concentration deviation, calculated in the same way as the initial risk deviation, but based on simulated adjusted data. C sync This represents the cooperative interference coefficient corresponding to the candidate scheme, read directly from the system cache, and is a real number greater than zero. D tfe This represents the distribution density of time-dimensional associated features in the integrated dataset within the candidate solution path. It is calculated using the kernel density estimation method and is expressed as the number of feature values ​​per unit path length.

[0066] The mathematical meaning of this formula is that the degree of intervention impact is directly proportional to the product of the corrected risk deviation and the collaborative interference coefficient, and inversely proportional to the distribution density of time-related features. This means that the higher the expected risk deviation after the implementation of the plan, or the greater the collaborative interference during the switching process, the greater the degree of intervention impact; conversely, the denser the distribution of time-related features along the plan path, the smaller the degree of intervention impact. A higher distribution density indicates that the data features of that path segment are concentrated, which may mean that the system has higher stability or redundancy in that area, thus better absorbing the disturbances brought by the intervention, and therefore the intervention impact is relatively small.

[0067] The entire calculation process is fully automated and executed by the system's algorithm engine. The engine first iterates through all generated candidate intervention classification schemes, performing the simulation, recalculation, and density analysis steps described above for each scheme in sequence. All intermediate results, including corrected risk deviation values, co-interference coefficients, and distribution densities, are temporarily stored in an in-memory database for later use in formula calculations. Finally, the calculated intervention impact value for each candidate scheme is appended to its metadata and passed to the intervention scheme generation module for further processing. This standardized and quantitative approach ensures objective comparison and evaluation of the impacts of different intervention schemes.

[0068] Example 4: This example involves the operation flow of the dynamic update module. This module updates the system database according to the optimal intervention classification scheme determined by the intervention scheme generation module and generates corresponding detection log information. The specific implementation details of this process are illustrated below with a concrete example.

[0069] Suppose the system is currently processing test data from a pregnant woman, with test ID GDM_2023_10_05_001. System analysis determines the optimal intervention category to be the dietary control plan with ID INTV_D_003. The dynamic update module then initiates the update process.

[0070] The update process first targets the detection and analysis paths in the database. These paths are data entities stored in a graph structure, where nodes represent historical and current detection time points and their corresponding serum biochemical indicator values, and edges represent the changes in indicator values ​​over time. The current state of this path records all detection data up to October 5, 2023. The update operations include: adding a new node corresponding to the latest detection time point (10:00 AM on October 5, 2023) and storing the predicted indicator values ​​after implementing the INTV_D_003 protocol (e.g., predicted blood glucose concentration 4.8 mmol / L, insulin level 18.2 pmol / L, glycated hemoglobin 5.9%); creating a new edge connecting this node to the node at the previous detection time point (September 7, 2023), with the edge attribute marked as "Intervention INTV_D_003 applied"; and modifying the attributes of an existing path, marking its current state as "Intervention implemented" and associating the intervention protocol ID with the path metadata.

[0071] While updating the detection and analysis path, the dynamic update module synchronously records the mapping relationship between historical detection time series and the serum biochemistry dataset for the current detection period. This mapping is achieved by creating a relational table in a relational database. For example, a mapping record is created with the primary key Map_001, linking the historical detection time series ID (TS_Seq_2023_07_to_09) with the serum biochemistry dataset ID for the current detection period (SS_Set_2023_10_05). The mapping relationship includes timestamp information, such as establishing a time-related link between the detection points on July 10, August 8, and September 7, 2023 in the historical series and the current detection point on October 5, 2023, and storing the differences in indicator values ​​at each time point (Δglucose concentration, Δinsulin level, etc.). This mapping data is used for subsequent analysis of the correlation between historical trends and the current intervention effect.

[0072] Based on the update results, the system generates detection log information. This log records detailed information about the update operation in a structured format. The log content includes: detection timestamp (recording the log generation time, such as 2023-10-05 14:30:00); the optimal intervention classification scheme used (storing the scheme ID INTV_D_003 and its complete parameters, such as carbohydrate intake ratio of 55%, 5 meals per day, etc.); a snapshot of the serum biochemistry dataset before the update (storing the raw data of the last test before the update in JSON format); a snapshot of the integrated dataset before the update (storing the time-dimensional correlation feature values ​​calculated before the update); and a snapshot of the physiological parameter information before the update (recording the body mass index of 23.1, gestational age of 28 weeks, and basal metabolic rate of 1550 kcal / day before the update), see Table 1.

[0073] Table 1: Detection Log Information Record Table.

[0074]

[0075] The entire update operation is completed within a single database transaction to ensure data consistency. At the start of the transaction, a database lock is acquired to prevent concurrent modifications. After the update is complete, the transaction is committed, the lock is released, and log information is written to a dedicated log database table. Log storage uses a partitioned table format, partitioned monthly by detection timestamp to optimize query performance.

[0076] The mapping relationships are stored in a dedicated table, which includes fields such as mapping ID, historical sequence ID, current dataset ID, mapping creation time, and time association details. The system periodically performs integrity checks on the mapping data to verify the validity of the association between historical sequences and the current dataset.

[0077] This implementation method ensures real-time updates to the system database, fully recording changes in status before and after intervention, and providing a traceable information foundation for subsequent data analysis. All operations are executed automatically through the system's backend service, requiring no manual intervention, and update results can be queried and approved through the management interface.

[0078] Example 5: See Figure 4 This embodiment involves the configuration and operation of the baseline state storage module and the intervention program library, as well as their collaborative work with the multidimensional indicator integration module. The baseline state storage module serves as the system's reference source, pre-storing baseline numerical ranges of physiological parameter information. The baseline body mass index (BMI) range is set at 18.5 to 24.9 kg / m², determined based on guidelines for the health of pregnant women, and divided into three sub-intervals: 18.5-22.9 is the ideal range, and 23.0-24.9 is the acceptable range. The baseline gestational age intervals are divided according to gestational stage: the first gestational age interval is 0-12 weeks, the second gestational age interval is 13-27 weeks, and the third gestational age interval is 28-40 weeks. A standard gestational age growth curve is set within each interval, including the average weekly weight gain range and physiological change parameters. The baseline metabolic rate threshold is dynamically adjusted according to the gestational age interval and BMI range, storing the minimum threshold of 1200 kcal / day and the maximum threshold of 1800 kcal / day for resting metabolic rate, and including a table of basal metabolic rate adjustment coefficients for different gestational ages. This baseline data is stored in a structured format in a dedicated table within a relational database. The table structure includes fields such as parameter type, lower limit, upper limit, applicable gestational age range, and version number. This baseline data is loaded into a memory cache during system initialization to improve access speed.

[0079] The intervention protocol library stores various intervention protocols and their implementation parameters corresponding to different risk levels. Risk levels are divided into three levels: Level 1 risk corresponds to mild abnormalities, Level 2 risk corresponds to moderate abnormalities, and Level 3 risk corresponds to severe abnormalities. Dietary control protocols include detailed parameters for each risk level. Level 1 risk dietary protocols include a daily carbohydrate intake of 50% to 55%, a protein intake of 20%, a fat intake of 25% to 30%, 5 meals per day, and a calorie restriction of 100 to 200 kcal per day. Level 2 risk dietary protocols adjust the carbohydrate intake to 45% to 50%, increase the protein intake to 22%, 6 meals per day, and a calorie restriction of 200 to 300 kcal per day. Level 3 risk dietary protocols further restrict carbohydrates to 40% to 45%, protein to 25%, 6 to 7 meals per day, and a calorie restriction of 300 to 400 kcal per day. All protocols specify the distribution of food types and meal intervals.

[0080] Exercise prescriptions are configured according to risk level and gestational week range. Level 1 risk exercise includes 30 minutes of walking daily, 5 times a week, maintaining a heart rate of 110-120 beats per minute; and prenatal yoga 3 times a week for 20 minutes each time. Level 2 risk exercise is adjusted to 20 minutes of walking daily, 6 times a week, maintaining a heart rate of 100-110 beats per minute; and water exercise 2 times a week for 25 minutes each time. Level 3 risk exercise includes 15 minutes of light walking daily, 7 times a week, maintaining a heart rate of 90-100 beats per minute; and breathing exercises 2 times a day for 10 minutes each time. Each plan includes exercise intensity level, duration, frequency, and precautions.

[0081] The medication intervention plan stores detailed parameters for oral medications and insulin therapy. Oral medication plans include metformin dosage ranges from 500 to 2000 mg daily, divided dose regimens, starting dose, and increment rules. Insulin therapy plans include dosage calculation formulas for various insulin types, injection times, dose adjustment rules, and monitoring frequencies. The plan is dynamically adjusted based on blood glucose monitoring results, and stores insulin dose adjustment tables corresponding to different blood glucose levels.

[0082] When generating the integrated dataset, the multidimensional index integration module adds phase synchronization markers to the time-series blood glucose concentration and insulin level data. These markers are calculated by analyzing the changing patterns of blood glucose concentration and insulin levels over time. The system collects blood glucose and insulin values ​​at continuous monitoring time stamps and uses signal processing methods to calculate the phase difference between the two sequences. The calculation process first standardizes the two time-series data to eliminate dimensional differences. Then, a sliding window analysis is used, with a window size of 24 hours and a step size of 1 hour. Within each window, the instantaneous phase angle of the blood glucose and insulin sequences is calculated, and the degree of phase synchronization between the two sequences is evaluated. The phase synchronization markers are divided into three levels: fully synchronized, partially synchronized, and out of sync, each corresponding to a numerical score. These markers are added to the metadata of the integrated dataset for subsequent analysis of time-dimensional correlation features.

[0083] During system operation, the baseline state storage module provides a real-time query interface, and other modules can obtain baseline data through the application programming interface. The intervention protocol library provides protocol retrieval functionality, supporting combined queries by risk level, gestational age, and intervention type. All data access operations are logged in an audit log to ensure data integrity. The phase synchronization marker calculation of the multidimensional indicator integration module is fully automated; the calculation process is triggered whenever new test data is generated, and the results are updated to the integrated dataset in real time. This entire implementation ensures that the system can perform consistent and reliable data analysis and intervention generation based on standardized baselines and a predefined protocol library.

[0084] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0085] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A system for analyzing data from a serum marker test for gestational diabetes, the system comprising: a database of serum marker test data; a data analysis module; and a user interface for inputting and outputting data to and from the data analysis module. The system comprises: a serum biochemical data acquisition module, which acquires serum samples of pregnant women through detection equipment, and constructs a serum biochemical data set containing blood glucose concentration, insulin level and glycated hemoglobin index; a multi-dimensional index integration module, which generates an integrated data set containing time dimension correlation characteristics based on the serum biochemical data set, in combination with historical detection time series and physiological parameter information in the current detection period; an abnormal risk analysis module, which processes the time dimension correlation characteristics in the integrated data set, and calculates a gestational diabetes risk deviation value based on the serum biochemical data set in the current detection period; a risk grade division module, which generates multiple candidate intervention classification schemes based on the serum biochemical data set when the gestational diabetes risk deviation value is abnormal, and calculates a synergistic interference coefficient when the physiological parameter information in the current detection period is switched to a reference state in combination with the integrated data set, and outputs the intervention influence degree of each candidate intervention classification scheme on the detection analysis process; an intervention scheme generation module, which screens the candidate intervention classification scheme with the lowest intervention influence degree, and generates the best intervention classification scheme in the current detection period.

2. The system for analysis of data of serum markers of gestational diabetes mellitus according to claim 1, characterized in that, The serum biochemical data set contains blood glucose concentration, insulin level and glycated hemoglobin values corresponding to continuous detection time stamps; The historical detection time series includes fluctuation trends of each serum biochemical index in the past three detection periods; The physiological parameter information includes body mass index, gestational age and basal metabolic rate in the current detection period.

3. The system for analysis of data of serum markers of gestational diabetes mellitus according to claim 2, characterized in that, The time dimension correlation characteristics of the integrated data set include time-varying correlation strength of blood glucose concentration and insulin level, and offset amount of glycated hemoglobin and historical fluctuation trend; The calculation method of the gestational diabetes risk deviation value is to extract the body mass index offset amount in the physiological parameter information in the current detection period, and perform weighted fusion in combination with the cumulative deviation amount of blood glucose concentration in the serum biochemical data set and the standard threshold value.

4. The system for analysis of data of markers of gestational diabetes serum detection according to claim 3, characterized in that, When the risk grade division module generates the candidate intervention classification scheme, it identifies abnormal boundary nodes of each index in the serum biochemical data set, marks the alert interval corresponding to each abnormal boundary node in the detection analysis path, and records the safe connection path of the current detection position and each alert interval as the candidate intervention area; The calculation basis of the synergistic interference coefficient is the proportional relationship between the total number of operation instructions required for switching the physiological parameter information in the historical record to the reference state and the number of operation instructions in the current switching process.

5. The system for analysis of data of serum markers of gestational diabetes mellitus according to claim 4, characterized in that, The acquisition method of the intervention influence degree includes calculating the gestational diabetes risk deviation value corresponding to a specific candidate intervention classification scheme, and combining the synergistic interference coefficient and the time dimension correlation characteristic distribution density of the integrated data set in the scheme path.

6. The system for analysis of data of serum markers of gestational diabetes mellitus according to claim 1, characterized in that, Further comprising: a dynamic update module, which updates the detection analysis path in the database in real time according to the best intervention classification scheme, and generates detection log information based on the update result; When the dynamic update module updates the detection analysis path, it synchronously records the mapping relationship between the historical detection time series and the serum biochemical data set in the current detection period.

7. The system for analysis of data of serum markers of gestational diabetes mellitus according to claim 6, characterized in that, The detection log information contains detection time stamp, adopted best intervention classification scheme, serum biochemical data set before update, integrated data set and physiological parameter information snapshot.

8. The system for analysis of data of markers of gestational diabetes mellitus in serum according to claim 1, characterized in that, The system further comprises a reference state storage module, which pre-stores a reference body mass index range, a reference gestational age interval and a reference metabolic rate threshold of the physiological parameter information.

9. The system for analysis of data of serum markers of gestational diabetes mellitus according to claim 8, characterized in that, The system further comprises an intervention scheme library, which stores execution parameters of a diet regulation scheme, an exercise prescription scheme and a drug intervention scheme corresponding to different risk levels.

10. The system for the analysis of data of markers of gestational diabetes serum detection according to claim 1, characterized in that, The multi-dimensional index integration module appends a phase synchronization mark of the time series blood glucose concentration data and the insulin level data when generating the integrated dataset.

Images