Intelligent management analysis and early warning system for ambulatory blood pressure data

Abstract

The invention relates to the technical field of intelligent medical treatment and health information, and discloses an intelligent management analysis and early warning system for ambulatory blood pressure data, which comprises a data acquisition and preprocessing module, a real-time monitoring and dynamic early warning module, a comparison and visual analysis module and an automatic report generation module, through the multi-source heterogeneous self-adaptive access and standardization technology of ambulatory blood pressure data, mainstream and novel blood pressure monitoring equipment are compatible, a new equipment data format can be autonomously learned and analyzed, meanwhile, through a data quality verification and conflict resolution mechanism, high accuracy of fused data is ensured, and a reliable foundation is laid for subsequent analysis.

Description

Technical Field

[0001] This invention relates to the field of smart healthcare and health information technology, and more specifically discloses an intelligent management, analysis and early warning system for dynamic blood pressure data. Background Technology

[0002] With the deep integration of modern medical technology with artificial intelligence, the Internet of Things and big data analysis, intelligent management, analysis and early warning systems for blood pressure data have emerged, providing important data support and decision-making assistance for the prevention and intervention of cardiovascular diseases.

[0003] The prior art patent document CN120531355A discloses a "method and system for abnormal early warning of continuous blood pressure measurement based on AI data analysis". This method includes: collecting pulse wave signals, triaxial accelerometer data, ambient light intensity signals, and skin conductance signals through sensors on a pre-set wearable device; generating multi-scale physiological characteristics; calculating systolic and diastolic blood pressure, and calculating vascular stiffness; calculating the risk level of aortic sclerosis; and alerting the user through the wearable device when the risk level of aortic sclerosis is higher than a preset risk level. According to the present invention, by simultaneously collecting and analyzing pulse wave, triaxial accelerometer, ambient light intensity, and skin conductance signals, the dynamic trends of systolic and diastolic blood pressure and vascular stiffness can be obtained, accurately quantifying the risk level of aortic sclerosis. While the aforementioned technical solutions can simultaneously collect pulse waves, triaxial accelerometer data, ambient light intensity, and skin conductance signals, and calculate systolic blood pressure, diastolic blood pressure, and vascular stiffness through multi-scale physiological characteristics, quantifying the risk of aortic sclerosis and forming a closed-loop management system of physiological parameters, vascular sclerosis, and risk decision-making to analyze and predict future blood pressure trends, existing technologies have limitations: First, data access compatibility is poor, relying heavily on specific brands or types of monitoring equipment and lacking the ability to learn new equipment data formats independently; new equipment access requires additional development of adaptation modules. Second, the early warning mechanism is not precise enough, failing to dynamically adjust early warning standards based on individual patient history and physiological rhythms, lacking auxiliary verification with multiple parameters such as heart rate, easily leading to false alarms or missed alarms, and lacking automated follow-up processing and tracking linkage after early warning. Third, data value mining is insufficient, lacking in-depth comparative analysis across time periods and complication subgroups, and failing to incorporate statistical testing methods to verify the clinical significance of data differences. Fourth, the entire process management is fragmented; report distribution requires manual intervention, core rule adjustments depend on technical personnel, and medical staff cannot flexibly configure the system according to clinical or research needs, making it difficult to form a closed-loop management system. Summary of the Invention

[0004] The main technical problem solved by this invention is to provide an intelligent management, analysis and early warning system for dynamic blood pressure data, which can solve the problems mentioned in the background art.

[0005] To solve the above-mentioned technical problems, according to one aspect of the present invention, more specifically, a dynamic blood pressure data intelligent management, analysis and early warning system, comprising: a data acquisition and preprocessing module, which acquires raw blood pressure data and corresponding blood pressure correlation parameters under different conditions, wherein the blood pressure correlation parameters include physiological correlation parameters and scenario parameters, and performs preprocessing on the raw blood pressure data and blood pressure correlation parameters to form blood pressure data and blood pressure correlation parameters in a unified format; The real-time monitoring and dynamic early warning module receives the blood pressure data in the unified format, constructs and updates the patient's personal blood pressure baseline, performs multi-dimensional correlation verification by combining physiological correlation parameters and scenario parameters, analyzes the personal blood pressure baseline and sets early warning judgment criteria, and outputs the abnormality level and abnormality type and triggers an early warning when the abnormal judgment conditions are met. The comparison and visualization analysis module calls the data output by the real-time monitoring and dynamic early warning module, supports individual time-series trend analysis, group comparison analysis and treatment efficacy statistical analysis, and presents the analysis results in an interactive manner. The automated report generation module calls upon the analysis results and processed data to generate a report that conforms to a preset structure and outputs it externally.

[0006] Furthermore, the data acquisition and preprocessing module includes: The data acquisition module supports Bluetooth, Wi-Fi, USB and API interface connections, and is compatible with medical and home blood pressure monitoring devices in scenarios such as hospitals, homes and health check centers to collect raw blood pressure data; The data preprocessing module performs outlier removal and missing value completion on the collected raw data, and conducts data quality assessment on the cleaned data. The data fusion and standardization module uses an adaptive parsing algorithm based on a rule engine to map multi-source heterogeneous data to a unified data model and completes data standardization processing through field alignment.

[0007] Furthermore, the real-time monitoring and dynamic early warning module includes: Personal baseline construction module: Receives the blood pressure data in the unified format, calculates and forms the patient's historical blood pressure baseline using the weighted moving average method; for new patients with insufficient historical blood pressure data, it calls the temporary baseline of the same characteristic group, and at the same time periodically receives new blood pressure data and blood pressure correlation parameters from the data acquisition and preprocessing module to update the personal blood pressure baseline data. Evaluation module: Receives personal baseline data, performs hierarchical processing on the data including correlation verification with physiological and scenario parameters, temporal feature extraction, scenario adaptive weighting, and anomaly decision-making, and outputs anomaly risk score and anomaly type.

[0008] Furthermore, the comparison and visualization analysis module includes: The individual time series analysis module, based on the ECharts or D3.js visualization library, calls the data output by the real-time monitoring and dynamic early warning module to perform time segmentation statistics on the patient's blood pressure data at different time periods, generate blood pressure trend curves, and calculate the blood pressure target achievement rate. The population analysis module calls the data output by the real-time monitoring and dynamic early warning module, supports segmentation according to patient characteristics, and uses box plots to display blood pressure statistics of different subgroups, enabling parallel comparison of population blood pressure data; The efficacy statistical verification module calls the data output by the real-time monitoring and dynamic early warning module, uses the t-test algorithm to verify the difference in blood pressure before and after medication for individual patients, and uses the one-way ANOVA algorithm to compare the differences in the efficacy of different treatment plans in the population.

[0009] Furthermore, the automated report generation module includes: The template customization and population module allows you to build configurable XML report templates, define the report structure and data field mapping relationships, and populate the corresponding positions in the template with the data output from the comparison and visualization analysis module. The report generation module uses document conversion technology to convert the filled template into PDF or DOCX format, and completes the report format rendering and standardized typesetting. The delivery and feedback module supports integration with HIS / LIS clinical systems to push reports, enabling electronic reports to be pushed through patient apps, and recording report viewing status and feedback information.

[0010] Furthermore, it also includes a medical scenario rule configuration module, used to adjust and manage anomaly judgment rules, analysis rules, and report generation rules. This medical scenario rule configuration module includes: The rule parameter configuration module uses a graphical GUI interface, which allows medical staff to visually adjust blood pressure abnormality thresholds, analysis indicator parameters, and report generation rules. The testing and verification module sets up a sandbox testing environment, uses historical blood pressure data to simulate the effect of new rules, and generates a rule verification report. The rules management module enables rule version control and log auditing, supports full or canary deployment of rules, and records the history of rule adjustments.

[0011] Furthermore, the system employs the HTTPS protocol to ensure data transmission security.

[0012] Furthermore, the data preprocessing module includes: The data cleaning submodule is used to remove outliers and fill in missing values ​​in the collected blood pressure-related raw data to obtain cleaned blood pressure-related raw data. The first evaluation submodule is used to evaluate the data quality of the cleaned blood pressure-related raw data and determine the data evaluation value of the cleaned blood pressure-related raw data; compare the data evaluation value with a preset evaluation threshold, and complete the data cleaning of the blood pressure-related raw data when the evaluation value is greater than or equal to the preset evaluation threshold. The method for assessing the data quality of the cleaned raw blood pressure data and determining the data assessment value of the cleaned raw blood pressure data includes: The raw blood pressure data before and after data cleaning are compared to determine the data reliability coefficient of the raw blood pressure data after data cleaning. ; in, This represents the data reliability coefficient of the raw blood pressure-related data after data cleaning. Indicates the number of residual outliers; This indicates the total amount of data after data cleaning; Indicates the number of missing values ​​to be filled; Indicates the number of original missing values; This represents the coefficient of variation of the raw blood pressure data after cleaning; This represents the coefficient of variation of the raw blood pressure data before cleaning; To represent a very small constant, avoid having a denominator of 0; Calculate the data evaluation value of the cleaned blood pressure-related raw data based on the data reliability coefficient; ; in, The data evaluation value represents the raw blood pressure data after cleaning; Indicates the completeness score; Indicates the accuracy score; Indicates the trend consistency score; This indicates that the clinical efficacy is worth scoring; This represents the overall reward and punishment coefficient; , , , These represent the weighting coefficients, all of which are greater than 0 and have a sum of 1.

[0013] Furthermore, the evaluation module includes: The associated verification submodule is used for: Obtain the current systolic blood pressure, diastolic blood pressure, heart rate, and blood oxygen saturation data from the raw blood pressure data; Calculate the deviations of the current systolic blood pressure, diastolic blood pressure and the corresponding baseline for the time period, and the deviations of the current heart rate and the resting heart rate, respectively, to obtain the systolic blood pressure deviation, diastolic blood pressure deviation and heart rate deviation; The systolic blood pressure deviation, diastolic blood pressure deviation, heart rate deviation, and blood oxygen saturation are correlated and verified. When the verification is passed, the first target data is output. The correlation verification includes heart rate-systolic blood pressure correlation verification, heart rate variability-systolic blood pressure correlation verification, and blood oxygen saturation-systolic blood pressure correlation verification. The temporal feature extraction submodule is used for: The standard deviation of systolic and diastolic blood pressure in the target data within the window is calculated using the sliding window method to obtain the short-term fluctuation coefficient. Based on linear regression fitting, with time as the independent variable and blood pressure as the dependent variable, the trend slope of blood pressure in the target data is calculated. Calculate the baseline drift in the target data and count the frequency of abnormal blood pressure exceeding the baseline within a preset time period; Short-term volatility coefficient, trend slope, baseline drift, and anomaly frequency are used as secondary target data. The scene adaptive correction submodule is used for: Construct the basic weights for each dimension of the data in the first and second target data; Preset scenario risk coefficients and measurement status labels; construct a scenario correction function based on the scenario risk coefficients and measurement status labels; The base weights are dynamically adjusted based on the scenario correction function to obtain the corrected weights. The normalization submodule is used to normalize the first target data and the second target data. The second evaluation submodule is used for: The evaluation value is calculated based on the normalized first target data, the second target data, and the corrected weights. The blood pressure abnormality level and type are determined based on the assessment value, and an early warning is issued.

[0014] The beneficial effects of this invention's intelligent management, analysis, and early warning system for dynamic blood pressure data are as follows: Through multi-source heterogeneous adaptive access and standardization technology for dynamic blood pressure data, it is not only compatible with mainstream and new blood pressure monitoring devices, but also can autonomously learn and parse data formats from new devices. Simultaneously, through data quality verification and conflict resolution mechanisms, it ensures high accuracy of the fused data, laying a reliable foundation for subsequent analysis. Furthermore, the real-time monitoring and dynamic early warning engine based on stream processing technology, combined with individual weighted moving average baseline and heart rate-assisted verification, achieves personalized, low-false-report blood pressure anomaly early warning, and links with automated reporting and tracking mechanisms to form a closed-loop management system of early warning, analysis, processing, and follow-up. At the same time, the visualization analysis module supporting multi-dimensional comparisons, incorporating statistical tests and subgroup stratified analysis, can intuitively present individual blood pressure time-series trends and group differences, providing a scientific basis for clinical decision-making that combines visualization and statistical support. Moreover, through data-driven automated report generation and configurable rule engine technology, it achieves full-process automation from data collection to report distribution, while empowering medical staff to independently configure analysis rules and early warning thresholds, significantly improving the efficiency and flexibility of blood pressure management and powerfully promoting the precision and intelligent development of hypertension management. Attached Figure Description

[0015] The present invention will now be described in further detail with reference to the accompanying drawings and specific implementation methods.

[0016] Figure 1 This is a schematic diagram of the system module architecture. Detailed Implementation

[0017] The present invention will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the present application can be combined with each other.

[0018] According to one aspect of the invention, such as Figure 1 As shown, an intelligent management, analysis, and early warning system for dynamic blood pressure data is provided, including: a data acquisition and preprocessing module, which acquires raw blood pressure data and corresponding blood pressure correlation parameters under different conditions. The blood pressure correlation parameters include physiological correlation parameters and scenario parameters. The module preprocesses the raw blood pressure data and blood pressure correlation parameters to form blood pressure data and blood pressure correlation parameters in a unified format. This module includes: Data acquisition module: Supports Bluetooth, Wi-Fi, USB, and API interface connections, adapting to medical and home blood pressure monitoring devices in scenarios such as hospitals, homes, and health check centers. It collects raw blood pressure data and corresponding blood pressure-related parameters, including physiological and scenario-related parameters. Specifically, it actively scans for nearby blood pressure monitoring devices (such as Omron HEM-7211) via Bluetooth BLE 5.0 and Wi-Fi 802.11b / g / n protocols, obtains the device model, MAC address, and supported communication protocols, and loads the corresponding lightweight driver plugin (such as "OMRON_Plugin_V2.3" adapted for Omron devices), completing device connection without manual intervention. Meanwhile, for offline measurement scenarios using home devices (such as elderly users forgetting to connect to the network), it supports batch data transmission from local device storage (up to 1,000 historical data records per transmission). During transmission, the system automatically verifies data integrity. Records lacking "measurement timestamp" or "patient ID" will be marked as "to be completed" and a notification will be sent to the user's mobile app to ensure data traceability.

[0019] The data preprocessing module performs outlier removal and missing value completion on the collected raw blood pressure data and corresponding blood pressure-related parameters, and conducts data quality assessment on the cleaned data. Outlier removal adopts the clinically accepted 3σ principle. For data within a single device and single acquisition cycle, the mean and standard deviation are calculated first. Data exceeding the mean ± 3 times the standard deviation (e.g., systolic blood pressure <80 mmHg or >240 mmHg, diastolic blood pressure <50 mmHg or >120 mmHg) are identified as extreme outliers and directly removed, with the reason for removal recorded (e.g., "systolic blood pressure 55 mmHg, exceeding the clinically reasonable lower limit"). Missing value completion employs a hybrid strategy combining forward imputation and adjacent mean correction. If a data record is missing "systolic blood pressure" or "diastolic blood pressure," the corresponding value from the previous valid record is temporarily used to fill the missing value. Subsequently, the mean of several valid data points before and after the missing record is used to correct the imputed value, ensuring the completed data closely approximates the actual measurement level.

[0020] Data fusion and standardization module: It adopts an adaptive parsing algorithm based on a rule engine to map multi-source heterogeneous data to a unified data model and completes data standardization processing through field alignment; Specifically, the adaptive parsing algorithm consists of three levels of execution logic: feature recognition, rule matching, and autonomous learning. First, it extracts four key features from the input data stream: device brand identifier, separator, data packet length, and number of fields. Then, it calculates the cosine similarity between these features and the features already stored in the system's device format feature library. The similarity calculation formula is as follows: In the formula The feature vector of the input data stream, This refers to the feature vector of a certain device that is already stored in the system device format feature library. Input feature vector The 1 eigenvalue, To store feature vectors The One eigenvalue; The field mapping stage maps the parsed data to a unified data model (including fields such as "patient ID, systolic blood pressure, diastolic blood pressure, heart rate, measurement timestamp, measurement status, and device number") to ensure that the meaning of data fields from different devices is consistent. For data conflicts arising from multiple measurements taken by different devices at the same time (e.g., ±1 minute) for the same patient, a weighted voting method is used to calculate the final valid data. The formula for calculating the final value is as follows: In the formula For the first The weight of the device For the first The standardized data from the device measurements are stored in the InfluxDB time-series database using a dual primary key of "patient ID - measurement timestamp" and simultaneously synchronized to the Kafka message queue, providing a unified data source for subsequent real-time monitoring and analysis modules.

[0021] The real-time monitoring and dynamic early warning module receives blood pressure data in a unified format, constructs and updates the patient's personal blood pressure baseline, performs multi-dimensional correlation verification by combining physiological and scenario parameters, analyzes the personal blood pressure baseline, sets early warning judgment criteria, and outputs the abnormality level and type and triggers an early warning when the abnormality judgment conditions are met. This module includes: The personal baseline construction module uses a weighted moving average method to calculate the patient's historical blood pressure baseline. For new patients with insufficient historical blood pressure data, a temporary baseline for the same characteristic group is used. At the same time, it periodically receives new blood pressure data and blood pressure-related parameters from the data acquisition and preprocessing module to update the personal blood pressure baseline data. The use of the temporary baseline for the same characteristic group is a reference standard set for newly connected patients with less than 3 days of effective personal blood pressure data. Essentially, it selects a group of patients from the system's existing historical data who are the same age ± 5 years, gender, and have basically the same underlying diseases as the target patient. It calculates the blood pressure statistical interval for this group and uses this interval as a temporary assessment benchmark for the target patient (it is automatically replaced by the personal blood pressure baseline after the target patient has accumulated ≥ 7 days of data).

[0022] This technical solution addresses situations where data is too limited, leading to insufficient sample size and significant errors. For example, if only one systolic blood pressure measurement (e.g., 130) is taken as the baseline, a subsequent fluctuation to 132 would be misinterpreted as exceeding the baseline. A provisional baseline, however, covers the range of most individuals (e.g., 130-140), where 132 is a crucial difference, minimizing misinterpretation. Essentially, it uses the stability of group data to compensate for the instability of limited individual data. Weighted averages also cannot be used directly because they require a sufficient sample size. If an individual only has two days of data, the weighted average is merely the mean of those two days, resulting in a small sample size that fails to reflect the true blood pressure level. Provisional baselines, using group data from dozens or even hundreds of individuals, filter out individual errors and are more reliable.

[0023] Specifically, the calculation is based on the patient's effective blood pressure data over the past 7 days, using a weighted moving average method to calculate the individual's historical blood pressure baseline. More recent data has a higher weight (e.g., data from 7 days ago has a weight of 0.2, while data from the current day has a weight of 1.16), ensuring that the baseline reflects the patient's latest blood pressure status. The calculation formula is as follows: In the formula, For the first Average blood pressure over the days For the first The weight of the heavens; For new patients whose data collection and preprocessing module outputs valid data for less than 3 days, the system calls a temporary baseline for the same age (e.g., ±5 years), gender, and underlying disease (e.g., hypertension, diabetes) group (constructed based on historical valid data statistics of patients with the same characteristics within the system). After the patient accumulates ≥7 days of valid data, the system switches to the individual baseline. In addition, the baseline data is automatically updated every day, and expired data from 7 days ago is removed and new data added on the same day is included.

[0024] Evaluation module: Receives data output from the personal baseline update module, performs hierarchical processing on the data including correlation verification, temporal feature extraction, scene adaptive weighting, and anomaly decision-making, and evaluates the data, outputting anomaly risk score and anomaly type.

[0025] The comparison and visualization analysis module acquires preprocessed data, performs multi-dimensional comparative analysis, and presents the results in an interactive format. This module includes: The individual time series analysis module, based on the ECharts or D3.js visualization library, calls the data output by the real-time monitoring and dynamic early warning module to perform time segmentation statistics on the patient's blood pressure data at different time periods, generate blood pressure trend curves, and calculate the blood pressure target achievement rate. Specifically, the system calls upon standardized time-series data output from the data acquisition and preprocessing module, constructs a dual-axis trend chart through a visualization library, with the horizontal axis representing a continuous time axis (which can be freely scaled by minute, hour, and day), and the vertical axis corresponding to systolic and diastolic blood pressure values ​​respectively. Two reference lines are simultaneously overlaid: for example, the orange dashed line represents the individual baseline (taken from the dynamically updated baseline of the real-time monitoring and dynamic early warning module), and the green solid line represents the clinical target line (matched to the corresponding standard based on the patient's underlying disease characteristics). The time axis supports precise positioning; selecting any time period allows calculation of the average blood pressure, target achievement rate, and fluctuation range within that interval. Data points are distinguished by different colors to indicate measurement status (e.g., blue dots for resting state, yellow triangles for after activity). Clicking on any data point will pop up a details window, displaying related information such as measurement time, heart rate, measurement scenario, and device number. If there is an early warning record at that time point, the early warning level and processing remarks are displayed simultaneously, achieving vertical linkage from "data" to "early warning" to "source tracing." Meanwhile, for long-term monitoring data, a smooth trend curve can be generated through polynomial fitting to intuitively present the changes in blood pressure over weeks, months, and seasons, and help identify rhythmic abnormalities such as morning hypertension and failure of blood pressure to drop at night.

[0026] The population analysis module calls the data output by the real-time monitoring and dynamic early warning module, supports segmentation according to patient characteristics, and uses box plots to display blood pressure statistics of different subgroups, enabling parallel comparison of population blood pressure data; Specifically, a hierarchical dimension library is first constructed based on patients' basic information (such as age, complications, and medication type). Healthcare professionals can select a single or combined dimension (e.g., ≥60 years old + diabetes mellitus with hypertension + calcium channel blocker medication) through a graphical interface to split the population data. After splitting, subgroups with insufficient sample size are filtered to ensure statistical validity. The core uses box plots to display the blood pressure distribution characteristics of each subgroup, with outliers marked with prominent red dots, intuitively presenting the degree of dispersion and extreme value distribution within the group. The charts support interactive drill-down, and hovering over a box displays the core indicators of the subgroup, such as "sample size, mean, and standard deviation". Clicking on a box allows further viewing of a list of basic information of all patients in that subgroup and a time-series trend chart of representative individuals. For overall differences between multiple groups, one-way ANOVA is automatically used for testing. When the statistical significance criteria are met, the difference between groups is marked in the chart to help quickly identify patients who are more sensitive to treatment response or have poorer blood pressure control.

[0027] The efficacy statistical verification module calls the data output by the real-time monitoring and dynamic early warning module, uses the t-test algorithm to verify the difference in blood pressure before and after medication for individual patients, and uses the one-way ANOVA algorithm to compare the differences in the efficacy of different treatment regimens in the population. Specifically, for individual patients, effective blood pressure data from two periods—"before medication" and "after medication"—were selected (note: the pre-medication period consisted of effective resting blood pressure data for 3–7 consecutive days before medication, and the post-medication period consisted of effective resting blood pressure data for the same duration after regular medication administration, ensuring the consistency of paired data). A paired t-test was used to verify the significance of the difference between the two groups of data. The calculation formula is as follows: In the formula, This is a statistic for paired t-tests, used to quantify the deviation of the blood pressure difference before and after medication from the "clinically indifferent threshold." Ultimately, by comparing it with the critical value of the t-distribution, it is determined whether the difference is statistically significant. The standard deviation of the difference (reflecting the dispersion of the blood pressure difference before and after medication) The smaller the value, the more stable the patient's blood pressure response to treatment. The mean of the differences between the two sets of data is calculated as follows: d = blood pressure measurement after medication - blood pressure measurement before medication. If d is negative, it means that blood pressure has decreased after medication (treatment is effective). If d is positive, it means that blood pressure has increased after medication (treatment may be ineffective or there may be adverse reactions). The number of paired samples. This is a reference value for the clinically indifferent threshold range, usually 0, representing the overall average difference in blood pressure before and after medication as 0, minus... This is to emphasize that we are performing a hypothesis test, specifically testing whether d is significantly different from 0. If we do not subtract... The formula can also be used for calculation, but it loses the comparative significance of hypothesis testing. It only represents the standardized value of d and cannot be directly used to infer whether the difference is significant.

[0028] The automated report generation module calls upon analysis results and processed data to generate reports conforming to a preset structure and output them externally. This module includes: The template customization and population module allows you to build configurable XML report templates, define the report structure and data field mapping relationships, and populate the corresponding positions in the template with the data output from the comparison and visualization analysis module. Specifically, the XML template is pre-defined according to clinical reporting standards, comprising four sections: patient basic information area, blood pressure time-series trend area, statistical analysis conclusion area, and early warning record area. Each section binds to the data source through a mapping relationship between field labels and data sources. For example, the "age" and "complication type" labels in the patient basic information area map to standardized patient information output by the data acquisition and preprocessing module; the trend chart label in the blood pressure time-series trend area calls the trend chart generated by the individual time-series analysis module in the comparison and visualization analysis module; and the pre- and post-medication value labels in the statistical analysis conclusion area are associated with the t-test results of the efficacy statistical verification module. It also supports department-level template customization. For example, the cardiology template adds a "cardiovascular risk level" field to the "statistical analysis conclusion area" to meet the reporting focus needs of different clinical scenarios.

[0029] The report generation module uses document conversion technology to convert the filled template into PDF or DOCX format, completing report format rendering and standardized typesetting. It uses the Apache POI toolkit to convert XML to DOCX format and the iText toolkit to generate PDF format, loading the system's preset typesetting style library during the conversion process. At the same time, it automatically generates a unique report number on the first page of the report (e.g., the rule is "Patient ID-Generation Timestamp-Report Type Code") and adds report descriptions (e.g., indicating data source, statistical method, and generation basis) on the last page, improving the standardization and traceability of the report.

[0030] The delivery and feedback module supports integration with HIS / LIS clinical systems to push reports, enabling electronic reports to be pushed through patient apps and recording report viewing status and feedback information. Specifically, it interfaces with hospital HIS / LIS systems via the HL7FHIR standard interface, carrying three key parameters when pushing reports: "unique report number, patient visit ID, and report generation time." This ensures that the clinical system can automatically link the report to the corresponding patient's medical record, allowing doctors to access the report synchronously by opening the patient's medical record without manual uploading. Simultaneously, the system records the push terminal (e.g., HIS / LIS / APP), push time, viewing time, viewing personnel, and feedback content for each report. If a patient does not view the report on the APP within a certain period (e.g., 72 hours), the system triggers a secondary reminder message on the APP (e.g., "You have a blood pressure management report to view. Click to learn more about your blood pressure"). Medical staff can view the report reach rate in real time through the log table. For high-risk patients who have not viewed the report (e.g., those with recent warning records), telephone follow-up reminders can be made, forming a closed loop of the entire process of "report generation - push - tracking - feedback".

[0031] It also includes a medical scenario rule configuration module, used to adjust and manage anomaly detection rules, analysis rules, and report generation rules. This module includes: The rule parameter configuration module uses a graphical GUI interface, allowing medical staff to visually adjust blood pressure abnormality thresholds, analysis indicator parameters, and report generation rules. For blood pressure abnormality threshold adjustment, the interface provides an "adjustment basis selection box," allowing medical staff to determine the reference standard for threshold adjustment without entering code. For example, after selecting the "diabetic and hypertension patients" group, the system loads the default threshold range for this group, and then fine-tunes the upper and lower limits of the threshold through the numerical input box. Furthermore, for report generation rules, the interface provides a "template field selection function," allowing medical staff to select whether to display fields such as "heart rate auxiliary verification results" and "historical baseline comparison chart" according to clinical needs. After selection, the system automatically updates the field mapping relationship of the report template, without requiring technical personnel to modify the template code.

[0032] The testing and validation module sets up a sandbox testing environment to simulate the effect of new rules using historical blood pressure data and generates a rule validation report. Specifically, the sandbox testing environment automatically extracts historical blood pressure data from the system over the past few months (e.g., 3 months) (including patient data from different age groups and complication types) to ensure that the simulation scenarios cover common clinical situations. After medical staff submit the new rule configuration, the system runs the new rule in the sandbox environment based on the historical data. The differences in key indicators under the new and old rules are compared, and the validation report will clearly present core data such as "number of warning triggers, warning accuracy, false alarm rate, and consistency of analysis results" (e.g., noting that "under the new rule, the warning accuracy rate for patients with diabetes and hypertension has increased from 82% to 91%, and the false alarm rate has decreased from 18% to 9%). At the same time, it provides adjustment suggestions for abnormal indicators (e.g., increased false alarm rate) (e.g., "It is recommended to narrow the deviation range of blood pressure abnormality thresholds").

[0033] The rules management module implements rule version control and log auditing functions, supports full or gray-scale deployment of rules, and records the history of rule adjustments. Specifically, the version control function names rules according to the rule type, adjustment date, and version number format (e.g., "Blood Pressure Abnormal Threshold-20241020-V1.2"). Medical staff can view historical version details at any time and roll back to any version with one click, avoiding risks caused by rule adjustment errors. The log auditing function records complete information for each rule adjustment, including the name of the person making the adjustment, the adjustment time, a comparison of parameters before and after the adjustment, and the basis for the adjustment (e.g., cited guideline clauses, departmental approval opinions). The log is tamper-proof, meeting the requirements for medical data traceability and compliance. In terms of deployment, full deployment is used for general regulations uniformly implemented throughout the hospital (e.g., basic early warning thresholds based on the latest guideline updates). After clicking "Full Deployment", the system immediately synchronizes the new rules to all clinical terminals. Gray-scale deployment is used for newly added or high-risk rules (e.g., customized analysis rules for specific complication groups). It is possible to "deploy in a certain ward of the cardiology department first", and after running for a certain period of time without abnormalities, it can be expanded to other wards through interface operations.

[0034] In the above technical solution, the data preprocessing module performs outlier removal and missing value completion operations on the collected raw data, and conducts data quality assessment on the cleaned data, specifically including: The data cleaning submodule is used to remove outliers and fill in missing values ​​in the collected blood pressure-related raw data to obtain cleaned blood pressure-related raw data. The first evaluation submodule is used to evaluate the data quality of the cleaned blood pressure-related raw data and determine the data evaluation value of the cleaned blood pressure-related raw data; compare the data evaluation value with a preset evaluation threshold, and complete the data cleaning of the blood pressure-related raw data when the evaluation value is greater than or equal to the preset evaluation threshold. The method for assessing the data quality of the cleaned raw blood pressure data and determining the data assessment value of the cleaned raw blood pressure data includes: The raw blood pressure data before and after data cleaning are compared to determine the data reliability coefficient of the raw blood pressure data after data cleaning. ; in, This represents the data reliability coefficient of the raw blood pressure-related data after data cleaning. Indicates the number of residual outliers; This indicates the total amount of data after data cleaning; Indicates the number of missing values ​​to be filled; Indicates the number of original missing values; This represents the coefficient of variation of the raw blood pressure data after cleaning; This represents the coefficient of variation of the raw blood pressure data before cleaning; To represent a very small constant, avoid having a denominator of 0; Calculate the data evaluation value of the cleaned blood pressure-related raw data based on the data reliability coefficient; ; in, The data evaluation value represents the raw blood pressure data after cleaning; Indicates the completeness score; Indicates the accuracy score; Indicates the trend consistency score; This indicates that the clinical efficacy is worth scoring; This represents the overall reward and punishment coefficient; , , , These represent the weighting coefficients, all of which are greater than 0 and have a sum of 1.

[0035] In this embodiment, This indicates the number of missing values ​​filled, which is the actual number of original missing records filled during data cleaning using methods such as mean imputation, linear interpolation, and time-series trend imputation.

[0036] In this embodiment, This indicates the number of original missing values; that is, the number of unrecorded or invalid records in the original blood pressure data before data cleaning.

[0037] In this embodiment, The comprehensive reward and punishment coefficient is used to dynamically adjust the data quality assessment results and reflect the clinical value of special patterns in the data. 0.9≤K≤1.2.

[0038] In this embodiment, ; ; ; The slope of the linear regression of the original blood pressure data after cleaning; The slope of the linear regression of the raw blood pressure data before cleaning; ; This represents the systolic blood pressure of the i-th record; This represents the diastolic blood pressure of the i-th record; This represents the clinical value function of blood pressure.

[0039] In this embodiment, The specific values ​​are shown in Table 1: Table 1 In this embodiment, ; The working principle and beneficial effects of the above technical solution are as follows: Outlier removal and missing value completion are performed on the collected raw blood pressure data. Outliers may be caused by measurement errors, equipment malfunctions, etc., and if not processed, they will seriously affect the accuracy of subsequent data analysis. The presence of missing values ​​also leads to incomplete data, affecting the analysis results. Data quality assessment is then performed on the cleaned raw blood pressure data. This assessment not only considers the comparison before and after data cleaning, measuring the effectiveness of data cleaning by calculating the Data Reliability Coefficient (DRC), but also integrates multiple dimensions such as completeness score, accuracy score, trend consistency score, and clinical validity score to calculate a data evaluation value (C). This comprehensive evaluation method can more accurately reflect the quality of the cleaned data and helps determine whether the data is suitable for subsequent analysis and application.

[0040] In the above technical solution, the evaluation module includes: The correlation verification submodule is used to: obtain the current systolic blood pressure, diastolic blood pressure, heart rate and blood oxygen saturation data from the raw blood pressure data; calculate the deviation of the current systolic blood pressure and diastolic blood pressure from the baseline of the corresponding time period, and the deviation of the current heart rate from the resting heart rate, to obtain the systolic blood pressure deviation, diastolic blood pressure deviation and heart rate deviation; The systolic blood pressure deviation, diastolic blood pressure deviation, heart rate deviation, and blood oxygen saturation are correlated and verified. When the verification is passed, the first target data is output. The correlation verification includes heart rate-systolic blood pressure correlation verification, heart rate variability-systolic blood pressure correlation verification, and blood oxygen saturation-systolic blood pressure correlation verification. The time-series feature extraction submodule is used to: calculate the standard deviation of systolic and diastolic blood pressure in the target data within the window based on the sliding window method, and obtain the short-term fluctuation coefficient; Based on linear regression fitting, with time as the independent variable and blood pressure as the dependent variable, the trend slope of blood pressure in the target data is calculated. Calculate the baseline drift in the target data and count the frequency of abnormal blood pressure exceeding the baseline within a preset time period; Short-term volatility coefficient, trend slope, baseline drift, and anomaly frequency are used as secondary target data. The scene adaptive correction submodule is used to: construct the basic weights corresponding to each dimension of the first target data and the second target data; Preset scenario risk coefficients and measurement status labels; construct a scenario correction function based on the scenario risk coefficients and measurement status labels; The base weights are dynamically adjusted based on the scenario correction function to obtain the corrected weights. The normalization submodule is used to normalize the first target data and the second target data. The second evaluation submodule is used for: The evaluation value is calculated based on the normalized first target data, the second target data, and the corrected weights. The blood pressure abnormality level and type are determined based on the assessment value, and an early warning is issued.

[0041] In this embodiment, the blood pressure-related raw data includes raw blood pressure data and corresponding blood pressure-related parameters. The blood pressure-related parameters include physiological-related parameters and scenario parameters. The raw blood pressure data includes, for example, systolic blood pressure (SBP) / diastolic blood pressure (DBP). The physiological-related parameters include heart rate (HR), heart rate variability (HRV): high-frequency component (HF: 0.15–0.4 Hz), low-frequency component (LF: 0.04–0.15 Hz), and blood oxygen saturation (SpO2). The scenario parameters include: measurement status label (ST): -resting (0) / post-exercise (1) ) / After meals (2) / Emotional fluctuations (3) / After taking medication (4) (User annotation + device-assisted judgment); Scenario risk coefficient (SC): Hospital: Postoperative (1.2) / Medication adjustment (1.1) / Routine hospitalization (1.0) Family: Elderly living alone (1.3) / History of falls (1.4) / Ordinary users (1.0) Physical examination center: BMI≥28 (1.3) / Family history of hypertension (1.2) / Routine physical examination (1.0); Device credibility (DC): Medical device (1.0) / Certified home device (0.9) / Uncertified device (0.7).

[0042] In this embodiment, the correlation between heart rate (HR) and systolic blood pressure (SBP) is verified as follows: Physiological basis: Under normal physiological conditions, when SBP increases, HR should increase slightly (sympathetic nerve activation) or remain unchanged (vagus nerve compensation). If there is an increase in SBP and a sudden drop in HR (ΔHR < -10 beats / min), it does not conform to physiological laws. Verification formula: If (ΔSBP > 10 mmHg and ΔHR < -5 beats / min) or (ΔSBP < -10 mmHg and ΔHR > 15 beats / min), the data is marked as suspicious, triggering a second measurement (5 minutes apart), and the average of the two valid data is used in subsequent calculations. Heart rate variability (HRV) - systolic blood pressure (SBP) correlation verification: Physiological basis: During hypertension stress, the sympathetic nervous system is excited, and the LF / HF ratio should increase (1.5); during hypotension, the vagus nerve is inhibited, and the LF / HF ratio should decrease (<0.8); Verification formula: If (SBP>baseline +15mmHg and LF / HF<1.0) or (SBP<baseline -15mmHg and LF / HF>0.8), the measurement status is marked as abnormal, and the user is prompted to confirm whether there is emotional fluctuation / exercise, and resting data is re-collected; Blood oxygen saturation (SpO2) - systolic blood pressure (SBP) correlation verification: Physiological basis: When SBP < 80 mmHg, tissue perfusion is insufficient, and SpO2 is likely < 95% (hypoxia). If SBP < 80 mmHg and SpO2 > 97%, it may be due to equipment error. Verification formula: If (SBP < 80 mmHg and SpO2 > 97%) OR (SBP > 180 mmHg and SpO2 < 93%), the data is marked as contradictory, and the data from medical equipment is given priority (DC = 1.0). If it is a home device, three measurements are triggered.

[0043] In this embodiment, the standard deviation of systolic and diastolic blood pressure in the target data within the window is calculated based on the sliding window method to obtain the short-term fluctuation coefficient. ; in, express Short-term volatility coefficient; The number of measurements within the window; This represents the measured value of systolic blood pressure during the i-th measurement; diastolic blood pressure is calculated in the same way. This is the average measured value of systolic blood pressure.

[0044] Based on linear regression fitting, with time as the independent variable and blood pressure as the dependent variable, the slope of the blood pressure trend in the target data is calculated, i.e.: ;in, The slope representing the trend of blood pressure; This is the blood pressure value; For a point in time; This represents the number of data points. Calculate the baseline drift in the target data and count the frequency of abnormal blood pressure deviations from the baseline within a preset time period, i.e.: Baseline drift = (average value of the same period in the past 7 days - initial baseline) / max(|initial baseline|, 1) × 100%; AF = number of times the baseline exceeds ±10 mmHg within 24 hours; In this embodiment, a basic weight matrix is ​​constructed, in which the weight of SBP deviation (ΔSBP) is 0.3, the weight of DBP deviation (ΔDBP) is 0.2, the weight of HRV-LF / HF is 0.15, the weight of short-term volatility coefficient (SF) is 0.15, the weight of trend slope (TS) is 0.1, the weight of SpO2 is 0.05, and the weight of anomaly frequency (AF) is 0.05.

[0045] In this embodiment, a scene correction function f(SC,ST) is constructed based on the scene risk coefficient (SC) and the measurement state label (ST). Preset scene parameters: Measurement status labels (ST): resting (0) / after exercise (1) / after meal (2) / mood fluctuation (3) / after medication (4); Scenario Risk Factor (SC): Hospital: Postoperative (1.2) / Medication adjustment (1.1) / Routine hospitalization (1.0); Family: Elderly living alone (1.3) / History of falls (1.4) / Regular users (1.0); Health checkup center: BMI≥28 (1.3) / family history of hypertension (1.2) / routine physical examination (1.0); Device credibility (DC): Medical devices (1.0) / Certified home devices (0.9) / Uncertified devices (0.7).

[0046] The basic weights are dynamically adjusted as follows: Post-operative hospital scenario (SC=1.2): TS weight ×1.5, HRV-LF / HF weight ×0.8; Post-exercise scenario at home (ST=1): ΔSBP weight ×0.6, SF weight ×1.2; High-risk scenario in health check-up center (SC=1.3): ΔSBP weight ×1.3, AF weight ×1.2; The weights are adjusted based on the device credibility (DC). If DC=0.7, all parameter weights are ×0.8.

[0047] In this embodiment, the evaluation value is calculated based on the normalized first target data, the second target data, and the corrected weights, including: in, Indicates the evaluation value; This represents the corrected weight value for the i-th dimension; This represents the normalized value of the i-th dimension.

[0048] In this embodiment, the anomaly level and anomaly type are determined based on the evaluation value, and an early warning is issued, including: Abnormalities are classified based on the assessment value S: S < 0.2 is normal, no warning; 0.2 ≤ S < 0.4 is mild abnormality, Level 1 warning; 0.4 ≤ S < 0.7 is moderate abnormality, Level 2 warning; S ≥ 0.7 is severe abnormality, Level 3 warning. Reasoning for anomalies by matching decision trees with rules: If ΔSBP < -15 mmHg + TS_SBP < -1.5 mmHg + HR > resting HR 15 bpm + SpO2 < 95%, it is considered orthostatic hypotension; TS_SBP is the trend slope of systolic blood pressure; If ΔSBP>20mmHg+LF / HF>2.0+ST=3+SF_SBP<5mmHg, it is diagnosed as stress hypertension; ST is the measurement status label; SF_SBP is the short-term fluctuation coefficient of systolic blood pressure; If ΔSBP > 15 mmHg + BD > 10% + AF > 4 times + SC = 1.3, it is considered as progression of essential hypertension; BD is the baseline drift; AF is the frequency of abnormalities; SC is the scenario risk coefficient. If ΔSBP < -10 mmHg + ST = 4 + HRV-HF is normal, it is determined to be drug-induced hypotension; HRV-HF is the normal high-frequency component in heart rate variability.

[0049] The working principle and beneficial effects of the above technical solution are as follows: Correlation verification is performed on systolic blood pressure deviation, diastolic blood pressure deviation, heart rate deviation, and blood oxygen saturation, covering correlation verifications such as heart rate-systolic blood pressure, heart rate variability-systolic blood pressure, and blood oxygen saturation-systolic blood pressure. Through this multi-dimensional correlation verification, unreasonable situations in the data can be effectively identified. For example, when the correlation between heart rate and systolic blood pressure does not conform to physiological laws, it can be determined that the data may contain errors or anomalies. Only data that passes the verification is used as the primary target data output, which greatly improves the accuracy and reliability of subsequent data analysis. Temporal characteristics of blood pressure data are extracted using multiple methods. Short-term fluctuation coefficients are obtained by calculating the standard deviations of systolic and diastolic blood pressure based on the sliding window method, which can reflect the blood... The study examines short-term fluctuations in blood pressure; it utilizes linear regression to calculate the trend slope of blood pressure, enabling analysis of its changing trends over time; it calculates baseline drift and the frequency of statistical anomalies to further understand the stability and abnormalities of blood pressure data; it constructs basic weights and, based on preset scenario risk coefficients and measurement status labels, builds scenario correction functions to dynamically adjust the basic weights; different scenarios (such as after exercise, during rest, etc.) and measurement states (such as measuring equipment, measuring environment, etc.) may affect blood pressure data, and by dynamically adjusting the weights, the evaluation results can be better adapted to different actual scenarios; it calculates the evaluation value using normalized data and adjusted weights, and determines the blood pressure abnormality level and type based on the evaluation value.

[0050] Of course, the above description is not a limitation of the present invention, and the present invention is not limited to the examples given above. Any changes, modifications, additions or substitutions made by those skilled in the art within the scope of the present invention are also within the protection scope of the present invention.

Claims

1. An intelligent management, analysis, and early warning system for dynamic blood pressure data, characterized in that, include: The data acquisition and preprocessing module acquires raw blood pressure data under different conditions. The raw blood pressure data includes raw blood pressure data and corresponding blood pressure correlation parameters. The blood pressure correlation parameters include physiological correlation parameters and scenario parameters. The module preprocesses the raw blood pressure data to form blood pressure data and blood pressure correlation parameters in a unified format. The real-time monitoring and dynamic early warning module receives the blood pressure data in the unified format, constructs and updates the patient's personal blood pressure baseline, performs multi-dimensional correlation verification by combining physiological correlation parameters and scenario parameters, analyzes the personal blood pressure baseline and sets early warning judgment criteria, and outputs the abnormality level and abnormality type and triggers an early warning when the abnormal judgment conditions are met. The comparison and visualization analysis module calls the data output by the real-time monitoring and dynamic early warning module, supports individual time-series trend analysis, group comparison analysis and treatment efficacy statistical analysis, and presents the analysis results in an interactive manner. The automated report generation module calls upon the analysis results and processed data to generate a report that conforms to a preset structure and outputs it externally.

2. The intelligent management, analysis, and early warning system for dynamic blood pressure data according to claim 1, characterized in that: The data acquisition and preprocessing module includes: The data acquisition module supports Bluetooth, Wi-Fi, USB and API interface connections, and is compatible with medical and home blood pressure monitoring devices in scenarios such as hospitals, homes, and health check centers, for collecting raw blood pressure-related data; The data preprocessing module performs outlier removal and missing value completion operations on the collected blood pressure-related raw data, and conducts data quality assessment on the cleaned data. The data fusion and standardization module uses an adaptive parsing algorithm based on a rule engine to map multi-source heterogeneous data to a unified data model and completes data standardization processing through field alignment.

3. The intelligent management, analysis, and early warning system for dynamic blood pressure data according to claim 1, characterized in that: The real-time monitoring and dynamic early warning module includes: The personal baseline construction module receives the blood pressure data in the unified format, calculates and forms the patient's historical blood pressure baseline using the weighted moving average method, and calls the temporary baseline of the same characteristic group for new patients whose historical blood pressure data is insufficient. At the same time, it periodically receives new blood pressure data and blood pressure correlation parameters from the data acquisition and preprocessing module to update the personal blood pressure baseline data. The assessment module receives personal baseline data, performs hierarchical processing on the data including correlation verification with physiological and scenario parameters, temporal feature extraction, scenario adaptive weighting, and anomaly decision-making, and outputs anomaly risk scores and anomaly types.

4. The intelligent management, analysis, and early warning system for dynamic blood pressure data according to claim 1, characterized in that: The comparison and visualization analysis module includes: The individual time series analysis module, based on the ECharts or D3.js visualization library, calls the data output by the real-time monitoring and dynamic early warning module to perform time segmentation statistics on the patient's blood pressure data at different time periods, generate blood pressure trend curves, and calculate the blood pressure target achievement rate. The population analysis module calls the data output by the real-time monitoring and dynamic early warning module, supports segmentation according to patient characteristics, and uses box plots to display blood pressure statistics of different subgroups, enabling parallel comparison of population blood pressure data; The efficacy statistical verification module calls the data output by the real-time monitoring and dynamic early warning module, uses the t-test algorithm to verify the difference in blood pressure before and after medication for individual patients, and uses the one-way ANOVA algorithm to compare the differences in the efficacy of different treatment plans in the population.

5. The intelligent management, analysis, and early warning system for dynamic blood pressure data according to claim 1, characterized in that: The automated report generation module includes: The template customization and population module allows you to build configurable XML report templates, define the report structure and data field mapping relationships, and populate the corresponding positions in the template with the data output from the comparison and visualization analysis module. The report generation module uses document conversion technology to convert the filled template into PDF or DOCX format, and completes the report format rendering and standardized typesetting. The delivery and feedback module supports integration with HIS / LIS clinical systems to push reports, enabling electronic reports to be pushed through patient apps, and recording report viewing status and feedback information.

6. The intelligent management, analysis, and early warning system for dynamic blood pressure data according to claim 1, characterized in that: It also includes a medical scenario rule configuration module, used to adjust and manage anomaly detection rules, analysis rules, and report generation rules. The medical scenario rule configuration module includes: The rule parameter configuration module uses a graphical GUI interface, which allows medical staff to visually adjust blood pressure abnormality thresholds, analysis indicator parameters, and report generation rules. The testing and verification module sets up a sandbox testing environment, uses historical blood pressure data to simulate the effect of new rules, and generates a rule verification report. The rules management module enables rule version control and log auditing, supports full or canary deployment of rules, and records the history of rule adjustments.

7. The intelligent management, analysis, and early warning system for dynamic blood pressure data according to claim 1, characterized in that: The system uses the HTTPS protocol to ensure data transmission security.

8. The intelligent management, analysis, and early warning system for dynamic blood pressure data according to claim 2, characterized in that, The data preprocessing module includes: The data cleaning submodule is used to remove outliers and fill in missing values ​​in the collected blood pressure-related raw data to obtain cleaned blood pressure-related raw data. The first evaluation submodule is used to evaluate the data quality of the cleaned blood pressure-related raw data and determine the data evaluation value of the cleaned blood pressure-related raw data; compare the data evaluation value with a preset evaluation threshold, and complete the data cleaning of the blood pressure-related raw data when the evaluation value is greater than or equal to the preset evaluation threshold. The method for assessing the data quality of the cleaned raw blood pressure data and determining the data assessment value of the cleaned raw blood pressure data includes: The raw blood pressure data before and after data cleaning are compared to determine the data reliability coefficient of the raw blood pressure data after data cleaning. ; in, This represents the data reliability coefficient of the raw blood pressure-related data after data cleaning. Indicates the number of residual outliers; This indicates the total amount of data after data cleaning; Indicates the number of missing values ​​to be filled; Indicates the number of original missing values; This represents the coefficient of variation of the raw blood pressure data after cleaning; This represents the coefficient of variation of the raw blood pressure data before cleaning; To represent a very small constant, avoid having a denominator of 0; Calculate the data evaluation value of the cleaned blood pressure-related raw data based on the data reliability coefficient; ; in, The data evaluation value represents the raw blood pressure data after cleaning; Indicates the completeness score; Indicates the accuracy score; Indicates the trend consistency score; This indicates that the clinical efficacy is worth scoring; This represents the overall reward and punishment coefficient; , , , These represent the weighting coefficients, all of which are greater than 0 and have a sum of 1.

9. The intelligent management, analysis, and early warning system for dynamic blood pressure data according to claim 3, characterized in that, The evaluation module includes: The associated verification submodule is used for: Obtain the current systolic blood pressure, diastolic blood pressure, heart rate, and blood oxygen saturation data from the raw blood pressure data; Calculate the deviations of the current systolic blood pressure, diastolic blood pressure and the corresponding baseline for the time period, and the deviations of the current heart rate and the resting heart rate, respectively, to obtain the systolic blood pressure deviation, diastolic blood pressure deviation and heart rate deviation; The systolic blood pressure deviation, diastolic blood pressure deviation, heart rate deviation, and blood oxygen saturation are correlated and verified. When the verification is passed, the first target data is output. The correlation verification includes heart rate-systolic blood pressure correlation verification, heart rate variability-systolic blood pressure correlation verification, and blood oxygen saturation-systolic blood pressure correlation verification. The temporal feature extraction submodule is used for: The standard deviation of systolic and diastolic blood pressure in the target data within the window is calculated using the sliding window method to obtain the short-term fluctuation coefficient. Based on linear regression fitting, with time as the independent variable and blood pressure as the dependent variable, the trend slope of blood pressure in the target data is calculated. Calculate the baseline drift in the target data and count the frequency of abnormal blood pressure exceeding the baseline within a preset time period; Short-term volatility coefficient, trend slope, baseline drift, and anomaly frequency are used as secondary target data. The scene adaptive correction submodule is used for: Construct the basic weights for each dimension of the data in the first and second target data; Preset scenario risk coefficients and measurement status labels; construct a scenario correction function based on the scenario risk coefficients and measurement status labels; The base weights are dynamically adjusted based on the scenario correction function to obtain the corrected weights. The normalization submodule is used to normalize the first target data and the second target data. The second evaluation submodule is used for: The evaluation value is calculated based on the normalized first target data, the second target data, and the corrected weights. The blood pressure abnormality level and type are determined based on the assessment value, and an early warning is issued.

Images