Blood pressure continuous measurement prediction method, device, equipment and storage medium

Abstract

The present application relates to the field of deep learning technology, and discloses a method, device, equipment and storage medium for continuous blood pressure measurement prediction. The method includes: using a blood pressure measurement system to monitor the blood pressure of a target user according to multiple first blood pressure measurement sites, and performing spatiotemporal correlation node discrimination and node screening to obtain multiple second blood pressure measurement sites; constructing a variable structure dynamic Bayesian network model and performing blood pressure state prediction to obtain a first blood pressure state prediction strategy; performing node prior probability density estimation and influence weight adjustment to obtain a second blood pressure state prediction strategy; performing blood pressure state prediction strategy fusion to generate a target blood pressure state prediction strategy; performing blood pressure anomaly detection through an initial Volterra functional series model to obtain a target blood pressure anomaly detection result; performing model optimization and continuous blood pressure anomaly detection, and outputting continuous blood pressure anomaly detection results. The present application improves the accuracy of continuous blood pressure measurement prediction.

Description

Technical Field

[0001] The present application relates to the field of deep learning technology, and in particular to a method, device, equipment and storage medium for continuous blood pressure measurement prediction. Background Art

[0002] In the current medical and health field, blood pressure is an important indicator for assessing individual cardiovascular health status. Its accurate measurement and continuous monitoring are crucial for the prevention, diagnosis and treatment of hypertension and related cardiovascular diseases. However, traditional blood pressure measurement methods mostly rely on intermittent manual measurement or use automatic blood pressure monitoring equipment to measure at fixed time points. These methods often fail to capture the dynamic changes of blood pressure throughout the day, and it is difficult to monitor and predict potential abnormal changes in blood pressure in real time, limiting the comprehensive understanding of individual blood pressure conditions.

[0003] With the rapid development of wearable technology and smart medical devices, it has become possible to realize real-time continuous monitoring of individual blood pressure. However, since blood pressure is affected by a variety of internal and external factors, its dynamic changes show complex nonlinear characteristics, which puts higher requirements on the analysis and processing of blood pressure monitoring data. How to accurately predict blood pressure status from continuously collected blood pressure data, especially to timely identify abnormal blood pressure fluctuations, has become an urgent problem to be solved. In addition, there are significant differences in blood pressure responses between individuals and under different conditions for the same individual, which requires that the blood pressure prediction method can not only handle complex data characteristics, but also be highly personalized and adaptable. However, existing blood pressure monitoring and prediction methods often ignore the influence of individual specificity and environmental factors, and it is difficult to meet the needs of accurate prediction and personalized health management. Summary of the invention

[0004] The present application provides a blood pressure continuous measurement prediction method, device, equipment and storage medium for improving the accuracy of blood pressure continuous measurement prediction.

[0005] In a first aspect, the present application provides a method for continuous blood pressure measurement prediction, the method comprising:

[0006] The blood pressure of the target user is monitored respectively according to the plurality of first blood pressure measurement sites by the blood pressure measurement system, and a set of associated parameters of each first blood pressure measurement site is obtained, and the spatiotemporal associated node discrimination and node screening of the plurality of first blood pressure measurement sites are performed according to the associated parameter set by the K nearest neighbor algorithm to obtain a plurality of second blood pressure measurement sites;

[0007] Constructing a variable structure dynamic Bayesian network model according to the plurality of second blood pressure measurement sites, and predicting the blood pressure states of the plurality of second blood pressure measurement sites by using the variable structure dynamic Bayesian network model to obtain a first blood pressure state prediction strategy for each second blood pressure measurement site;

[0008] Based on the core attention mechanism, node prior probability density estimation is performed on each of the plurality of second blood pressure measurement sites to obtain a prior probability density of each second blood pressure measurement site, and the influence weight of the first blood pressure state prediction strategy is adjusted according to the prior probability density to obtain a second blood pressure state prediction strategy for each second blood pressure measurement site;

[0009] Based on a rectangular segmentation global optimization algorithm, a blood pressure state prediction strategy is integrated for each second blood pressure measurement site to generate a target blood pressure state prediction strategy for the blood pressure measurement system;

[0010] Based on the target blood pressure state prediction strategy, the blood pressure measurement system collects blood pressure data to obtain blood pressure time series data, and inputs the blood pressure time series data into a preset initial Volterra functional series model to perform blood pressure anomaly detection to obtain a target blood pressure anomaly detection result;

[0011] The initial Volterra functional series model is optimized according to the target blood pressure abnormality detection result by using an adaptive bat algorithm to obtain a target Volterra functional series model, and the blood pressure measurement system is continuously tested for blood pressure abnormality by using the target Volterra functional series model to output the continuous blood pressure abnormality detection result.

[0012] In a second aspect, the present application provides a blood pressure continuous measurement prediction device, the blood pressure continuous measurement prediction device comprising:

[0013] A monitoring module, configured to respectively monitor the blood pressure of a target user according to a plurality of first blood pressure measurement sites through a blood pressure measurement system, obtain a set of associated parameters for each first blood pressure measurement site, and perform spatiotemporal associated node discrimination and node screening on the plurality of first blood pressure measurement sites according to the associated parameter set through a K-nearest neighbor algorithm, to obtain a plurality of second blood pressure measurement sites;

[0014] A prediction module, configured to construct a variable structure dynamic Bayesian network model according to the plurality of second blood pressure measurement sites, and predict the blood pressure states of the plurality of second blood pressure measurement sites by using the variable structure dynamic Bayesian network model to obtain a first blood pressure state prediction strategy for each second blood pressure measurement site;

[0015] an adjustment module, configured to respectively estimate the node prior probability density of the plurality of second blood pressure measurement sites based on a core attention mechanism to obtain a prior probability density of each second blood pressure measurement site, and to adjust the influence weight of the first blood pressure state prediction strategy according to the prior probability density to obtain a second blood pressure state prediction strategy for each second blood pressure measurement site;

[0016] a fusion module, configured to fuse the second blood pressure state prediction strategy of each second blood pressure measurement site based on a rectangular segmentation global optimization algorithm to generate a target blood pressure state prediction strategy for the blood pressure measurement system;

[0017] A detection module, configured to collect blood pressure data from the blood pressure measurement system based on the target blood pressure state prediction strategy to obtain blood pressure time series data, and input the blood pressure time series data into a preset initial Volterra functional series model to perform blood pressure anomaly detection to obtain a target blood pressure anomaly detection result;

[0018] The output module is used to optimize the initial Volterra functional series model according to the target blood pressure abnormality detection result through an adaptive bat algorithm to obtain a target Volterra functional series model, and perform continuous blood pressure abnormality detection on the blood pressure measurement system through the target Volterra functional series model to output the continuous blood pressure abnormality detection result.

[0019] The third aspect of the present application provides a blood pressure continuous measurement prediction device, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the blood pressure continuous measurement prediction device executes the above-mentioned blood pressure continuous measurement prediction method.

[0020] A fourth aspect of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the above-mentioned blood pressure continuous measurement prediction method.

[0021] In the technical solution provided by the present application, by continuously monitoring and obtaining a set of associated parameters at multiple first blood pressure measurement sites, real-time and continuous tracking of the target user's blood pressure can be achieved. This not only helps to capture the dynamic changes of blood pressure in a short period of time, but also can promptly detect abnormal fluctuations in blood pressure. By constructing a variable structure dynamic Bayesian network model and combining the kernel attention mechanism to estimate the node prior probability density, the blood pressure state of each second blood pressure measurement site can be more accurately predicted. The nonlinear characteristics and time dependence of blood pressure data can be effectively processed to improve the accuracy of blood pressure state prediction. Through the spatiotemporal correlation node discrimination and node screening based on the K nearest neighbor algorithm, and the fusion of blood pressure state prediction strategies based on the rectangular segmentation global optimization algorithm, the influence of individual differences and environmental factors can be fully considered to achieve the personalization and adaptability of blood pressure prediction. By inputting the blood pressure time series data into the preset initial Volterra functional series model for blood pressure anomaly detection, and using the adaptive bat algorithm for model optimization, it is not only possible to effectively identify abnormal fluctuations in blood pressure, but also to adaptively optimize the anomaly detection model to adapt to the complexity of blood pressure changes. Real-time monitoring, accurate prediction and effective anomaly detection of blood pressure are achieved, providing users with a comprehensive and efficient blood pressure health management solution. This not only helps users understand their blood pressure status in a timely manner, but also provides data support, thereby improving the accuracy of continuous blood pressure measurement predictions. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0023] Figure 1 This is a schematic diagram of an embodiment of the method for continuous blood pressure measurement prediction in the embodiment of the present application;

[0024] Figure 2 This is a schematic diagram of an embodiment of a blood pressure continuous measurement prediction device in an embodiment of the present application. DETAILED DESCRIPTION

[0025] The embodiments of the present application provide a method, device, equipment and storage medium for continuous measurement prediction of blood pressure. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described here can be implemented in an order other than that illustrated or described here. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0026] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 , an embodiment of the blood pressure continuous measurement prediction method in the embodiment of the present application includes:

[0027] Step S101: using a blood pressure measurement system to monitor the blood pressure of a target user according to a plurality of first blood pressure measurement sites, respectively, to obtain a set of associated parameters of each first blood pressure measurement site, and using a K-nearest neighbor algorithm to perform spatiotemporal associated node discrimination and node screening on the plurality of first blood pressure measurement sites according to the associated parameter set, to obtain a plurality of second blood pressure measurement sites;

[0028] It is understandable that the execution subject of the present application may be a blood pressure continuous measurement prediction device, or a terminal or a server, which is not limited here. The present application embodiment is described by taking a server as the execution subject as an example.

[0029] Specifically, the target user's blood pressure measurement site is initialized and set, and multiple first blood pressure measurement sites are determined. The blood pressure measurement system monitors the target user's blood pressure according to the multiple first blood pressure measurement sites. During the monitoring process, a set of associated parameters is obtained from each measurement site. These parameters not only include direct blood pressure values, but also include pulse wave velocity, blood flow, measurement time, user heart rate, user body temperature, and environmental parameters, etc. These data together constitute a multi-dimensional perspective for blood pressure state analysis. The associated parameter set of each first blood pressure measurement site is analyzed by the K nearest neighbor algorithm. By evaluating the contribution rate of the parameters and sorting the parameter features, a specific feature parameter set is obtained for each measurement site to determine which parameters are most critical for blood pressure state prediction. Through the dynamic spatiotemporal correlation weight function, the spatiotemporal correlation of all first blood pressure measurement sites is calculated according to the feature parameter set. This calculation takes into account the spatial distribution of each measurement site and also integrates the change trend in the time series, thereby generating a set containing spatiotemporal correlation data. This set reflects the mutual influence and correlation degree between different measurement sites in time and space, and provides an important basis for the comprehensive prediction of blood pressure state. Through the node screening function with complexity adjustment, all the first blood pressure measurement sites are screened according to the spatiotemporal correlation data set. The screening process is based on the spatiotemporal correlation of each measurement site and its importance to blood pressure prediction, aiming to screen out the most representative and predictive sites from the numerous first blood pressure measurement sites, i.e., multiple second blood pressure measurement sites.

[0030] Step S102: constructing a variable structure dynamic Bayesian network model according to the multiple second blood pressure measurement sites, and predicting the blood pressure states of the multiple second blood pressure measurement sites by using the variable structure dynamic Bayesian network model to obtain a first blood pressure state prediction strategy for each second blood pressure measurement site;

[0031] Specifically, dependency analysis is performed on multiple second blood pressure measurement sites to understand and determine the interactions and dependencies between the measurement sites, and a corresponding dynamic Bayesian network model is constructed. The initial model reflects the basic relationships between the blood pressure measurement sites and their impact on the prediction of blood pressure status. Due to the variability of physiological parameters and environmental conditions, these dependencies are not static, so a model that can adapt to such changes is required. The initial dynamic Bayesian network model is adjusted by a variable structure dynamic weight update function, and the adjustment of the node connection weight is based on the latest monitoring data and prediction results. This adjustment enables the Bayesian network model to have the ability to dynamically adapt, and can reflect the dynamic changes in the blood pressure status and the relationship between the measurement sites that change over time, and obtain a variable structure dynamic Bayesian network model. This model can not only more accurately describe the current characteristics of the blood pressure state, but also predict future trends. Through the variable structure dynamic Bayesian network model, the blood pressure status of multiple second blood pressure measurement sites is predicted. The model will comprehensively consider the specific data of each measurement site and the dynamic dependencies between them, and generate a first blood pressure state prediction strategy for each second blood pressure measurement site. This prediction strategy includes an assessment of the current blood pressure status and also predicts the possible changing trend of blood pressure in the future.

[0032] Step S103: performing node prior probability density estimation for each of the second blood pressure measurement sites based on the core attention mechanism to obtain a prior probability density of each second blood pressure measurement site, and adjusting the influence weight of the first blood pressure state prediction strategy according to the prior probability density to obtain a second blood pressure state prediction strategy for each second blood pressure measurement site;

[0033] Specifically, the characteristic parameter sets of multiple second blood pressure measurement sites are mapped to a high-dimensional space through the kernel function conversion formula. By calculating the similarity between any two points in the original data space and converting this similarity into a high-dimensional feature space through nonlinear mapping, the limitations of traditional linear models in processing complex data are overcome. The kernel function conversion process not only enhances the model's ability to capture the intrinsic structure of blood pressure measurement data, but also flexibly controls the sensitivity of similarity calculation by adjusting the kernel width parameter σ, thereby obtaining a refined high-dimensional feature representation for each second blood pressure measurement site. The attention weight is calculated for the high-dimensional feature space of each second blood pressure measurement site based on the kernel attention mechanism. The kernel attention mechanism assigns an attention weight to each feature by evaluating the contribution of different features to the prediction of blood pressure status, thereby highlighting those features that are most critical to the prediction. In this way, the model can automatically identify and focus on the most informative features, effectively improving the accuracy and efficiency of blood pressure status prediction. According to the attention feature weight, the prior probability density of each second blood pressure measurement site is calculated separately. By integrating the attention weighted features of each measurement site, its prior distribution in the prediction of blood pressure status is estimated, providing a quantitative prior knowledge basis for subsequent prediction strategies. Based on the prior probability density, the influence weight of the first blood pressure state prediction strategy is adjusted to obtain the second blood pressure state prediction strategy for each second blood pressure measurement site. This adjustment process is based on prior knowledge and combined with actual measurement data to dynamically adjust the parameters of the prediction model to more accurately reflect the changing trend of the blood pressure state.

[0034] Step S104, performing blood pressure state prediction strategy fusion on the second blood pressure state prediction strategy of each second blood pressure measurement site based on a rectangular segmentation global optimization algorithm to generate a target blood pressure state prediction strategy for the blood pressure measurement system;

[0035] Specifically, multiple second blood pressure measurement sites are segmented into rectangles to refine the spatial area of ​​blood pressure measurement so as to more accurately perform subsequent prediction error analysis and optimization. The process of rectangular segmentation is based on the spatial distribution characteristics of the blood pressure measurement sites, and these sites are divided into multiple rectangular areas. Each area represents a specific measurement space range. Such a division helps to capture and understand the differences in blood pressure changes at different spatial locations. A global search is performed on the rectangular segmentation area through a rectangular segmentation global optimization algorithm. The global optimization algorithm aims to find a parameter vector that minimizes the prediction error in each rectangular area, and ensures that the solution found is optimal in the entire parameter space through global search. Various factors affecting the prediction of blood pressure status, such as time, environmental conditions, and individual differences, are comprehensively considered to ensure that the minimization of prediction error is achieved in a global scope. According to the minimized prediction error parameter vector of each rectangular segmentation area, the blood pressure status prediction strategy is fused. The prediction strategies obtained in different rectangular areas are combined to form a unified target blood pressure status prediction strategy for the entire blood pressure measurement system. In the fusion process, the weights and importance of the prediction strategies of different areas need to be considered to ensure that the final target prediction strategy reflects both the global optimization results and the specific needs of the local area.

[0036] Step S105: collecting blood pressure data from the blood pressure measurement system based on the target blood pressure state prediction strategy to obtain blood pressure time series data, and inputting the blood pressure time series data into a preset initial Volterra functional series model to perform blood pressure anomaly detection to obtain a target blood pressure anomaly detection result;

[0037] Specifically, based on the target blood pressure state prediction strategy, the blood pressure measurement system collects blood pressure data, including real-time measurement of blood pressure, real-time recording of the measurement environment and user status, which together constitute the blood pressure time series data. The blood pressure time series data is input into the preset initial Volterra functional series model, which is a highly complex mathematical model that can handle nonlinear relationships. The model can finely analyze the dynamic changes of blood pressure data over time, and then extract the abnormal fluctuation period. The Volterra model can predict the possible change trend of blood pressure by considering the impact of blood pressure changes at different time points in the past on the current blood pressure value. This prediction is not only based on the data of a single time point, but also based on the overall analysis of data over a period of time, so as to more accurately capture the dynamic change characteristics of blood pressure. After obtaining the target abnormal fluctuation period, these periods are further tested for blood pressure anomalies by the anomaly detection judgment function. This judgment function is based on the difference between the predicted blood pressure value and the reference blood pressure value. By setting an abnormal detection threshold δ, it is judged whether the difference is large enough to determine whether it is abnormal. The advantage of this method is that it can not only identify obvious blood pressure abnormalities, but also flexibly adapt to the normal blood pressure fluctuation range of different users by adjusting the threshold, thereby improving the accuracy and individualization of abnormal detection.

[0038] Step S106: Using the adaptive bat algorithm, the initial Volterra functional series model is optimized according to the target blood pressure abnormality detection result to obtain the target Volterra functional series model, and the blood pressure measurement system is continuously tested for blood pressure abnormality using the target Volterra functional series model to output the continuous blood pressure abnormality detection result.

[0039] Specifically, the initial Volterra functional series model is optimized by the adaptive bat algorithm to improve the accuracy and efficiency of continuous blood pressure anomaly detection. As a global optimization algorithm, the adaptive bat algorithm can quickly locate the optimal solution in a complex search space. By using this algorithm, the parameters in the initial Volterra functional series model are optimized according to the obtained target blood pressure anomaly detection results, and a set of model parameters that can minimize the prediction error are found. The optimized model parameter solution is used to update the parameters in the initial Volterra functional series model, and the model is recalibrated according to the actual blood pressure anomaly detection results to ensure that the model can more accurately reflect the dynamic characteristics of blood pressure changes in future blood pressure anomaly detection. The parameter update is not only based on the optimization results in a statistical sense, but also improves the actual performance of the model, so that the model can better adapt to the changes in actual blood pressure data, thereby improving the accuracy and reliability of continuous blood pressure anomaly detection. Through the optimized target Volterra functional series model, the blood pressure measurement system performs blood pressure anomaly detection, and the output continuous blood pressure anomaly detection results will more accurately reflect the user's blood pressure status and possible abnormal fluctuations.

[0040] In the embodiment of the present application, by continuously monitoring and obtaining the associated parameter set at multiple first blood pressure measurement sites, real-time and continuous tracking of the target user's blood pressure can be achieved. This not only helps to capture the dynamic changes of blood pressure in a short period of time, but also can promptly detect abnormal fluctuations in blood pressure. By constructing a variable structure dynamic Bayesian network model and combining the kernel attention mechanism to estimate the node prior probability density, the blood pressure state of each second blood pressure measurement site can be more accurately predicted. The nonlinear characteristics and time dependence of blood pressure data can be effectively processed to improve the accuracy of blood pressure state prediction. Through the spatiotemporal correlation node discrimination and node screening based on the K nearest neighbor algorithm, and the blood pressure state prediction strategy fusion based on the rectangular segmentation global optimization algorithm, the influence of individual differences and environmental factors can be fully considered to achieve the personalization and adaptability of blood pressure prediction. By inputting the blood pressure time series data into the preset initial Volterra functional series model for blood pressure anomaly detection, and using the adaptive bat algorithm for model optimization, not only can the abnormal fluctuations of blood pressure be effectively identified, but also the abnormal detection model can be adaptively optimized to adapt to the complexity of blood pressure changes. Real-time monitoring, accurate prediction and effective abnormality detection of blood pressure are achieved, providing users with a comprehensive and efficient blood pressure health management solution. This not only helps users understand their blood pressure status in a timely manner, but also provides data support, thereby improving the accuracy of continuous blood pressure measurement predictions.

[0041] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0042] (1) Initializing the blood pressure measurement site for the target user to obtain a plurality of first blood pressure measurement sites;

[0043] (2) monitoring the blood pressure of the target user according to the plurality of first blood pressure measurement sites respectively through the blood pressure measurement system, and obtaining a set of associated parameters of each first blood pressure measurement site respectively, the set of associated parameters including: blood pressure value, pulse wave velocity, blood flow, measurement time, user heart rate, user body temperature and environmental parameters;

[0044] (3) evaluating the parameter contribution rate and ranking the parameter characteristics of the associated parameter set of each first blood pressure measurement site by using the K nearest neighbor algorithm to obtain a characteristic parameter set of each first blood pressure measurement site;

[0045] (4) calculating the spatiotemporal correlation of the plurality of first blood pressure measurement sites according to the characteristic parameter set by using a dynamic spatiotemporal correlation weight function, thereby obtaining a spatiotemporal correlation data set of the plurality of first blood pressure measurement sites;

[0046] (5) Using a complexity-adjusted node screening function, a plurality of first blood pressure measurement sites are screened according to a spatiotemporal correlation data set to obtain a plurality of second blood pressure measurement sites.

[0047] Specifically, the target user's blood pressure measurement site is initialized, and a suitable blood pressure measurement site, such as the wrist, upper arm or other body parts, is selected, and the specific positions of these measurement sites are accurately calibrated to ensure the accuracy and consistency of subsequent measurements. Through the blood pressure measurement system, the target user's blood pressure is monitored according to multiple first blood pressure measurement sites, and multiple associated parameters including blood pressure value, pulse wave velocity, blood flow, measurement time, user heart rate, user body temperature and environmental parameters are collected. These parameters can not only reflect the user's blood pressure state, but also provide important information about the user's health status and measurement environment. For example, for the same target user, the blood pressure values ​​measured at different ambient temperatures may be different. The accuracy and reliability of the blood pressure monitoring results are ensured by comprehensively analyzing these associated parameters. The parameter contribution rate and parameter feature sorting of the associated parameter set of each first blood pressure measurement site are evaluated by the K nearest neighbor algorithm. As a machine learning algorithm, the K nearest neighbor algorithm can identify the parameters that have the greatest influence on blood pressure prediction. Through algorithm analysis, a characteristic parameter set of each measurement site is obtained, which will guide the subsequent blood pressure state analysis and prediction. For example, the algorithm may find that under certain conditions, the change in the user's heart rate has a greater impact on the blood pressure value than other parameters, which helps to improve the accuracy of blood pressure prediction. Through the dynamic spatiotemporal correlation weight function, the spatiotemporal correlation of multiple first blood pressure measurement sites is calculated according to the characteristic parameter set. Through mathematical models and algorithms, the correlation of different measurement sites under different time and space conditions is analyzed and calculated to obtain a comprehensive spatiotemporal correlation data set. This analysis helps to understand the mutual influence between different measurement sites and the impact of time and space changes on blood pressure monitoring results. Through the node screening function with complexity adjustment, multiple first blood pressure measurement sites are node screened according to the spatiotemporal correlation data set to obtain multiple second blood pressure measurement sites. Based on the above analysis results, through mathematical models and algorithms, the most critical sites for blood pressure status prediction are screened from all first blood pressure measurement sites. For example, after knowing that the data of the upper arm blood pressure measurement site is more effective in predicting the overall blood pressure status, the system may use it as the second blood pressure measurement site to optimize and improve the overall effect of blood pressure monitoring and prediction.

[0048] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0049] (1) Analyze the dependency relationship of multiple second blood pressure measurement sites and construct a corresponding initial dynamic Bayesian network model;

[0050] (2) The node connection weights of the initial dynamic Bayesian network model are adjusted through the variable structure dynamic weight update function to obtain a variable structure dynamic Bayesian network model;

[0051] (3) Predicting the blood pressure status of multiple second blood pressure measurement sites using a variable structure dynamic Bayesian network model to obtain a first blood pressure status prediction strategy for each second blood pressure measurement site.

[0052] Specifically, the dependencies between the second blood pressure measurement sites are analyzed to understand the interactions and influences between different blood pressure measurement sites. By collecting and analyzing historical blood pressure data from these measurement sites, statistical methods or machine learning algorithms are used to reveal the correlations and dependencies between them. An initial dynamic Bayesian network model is constructed based on the dependencies. Dynamic Bayesian network is a tool for processing time series data that can simulate the probability relationship between variables that changes over time. In this model, each second blood pressure measurement site is regarded as a node, and the connections between nodes represent the dependencies between them. The construction of the initial model is based on the understanding of known data and dependencies, providing a basic framework for subsequent model optimization and blood pressure state prediction. The node connection weights of the initial dynamic Bayesian network model are adjusted by a variable structure dynamic weight update function. The connection weights between nodes are dynamically adjusted according to new data input to reflect the latest dependencies and interactions. For example, if it is found that the contribution of data from a certain blood pressure measurement site to the prediction of the overall blood pressure state increases, the connection weights of this site with other nodes will increase accordingly. This dynamic adjustment enables the Bayesian network model to adapt to data changes and maintain the accuracy and real-time nature of its predictive ability. The variable structure dynamic Bayesian network model is used to predict the blood pressure status of multiple second blood pressure measurement sites, and finally the first blood pressure status prediction strategy of each second blood pressure measurement site is obtained. The model comprehensively uses the weighted network structure and the latest blood pressure measurement data to predict the blood pressure status of each measurement site. This includes the current blood pressure value and may also predict the trend of blood pressure changes in the future.

[0053] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0054] (1) A kernel function conversion formula is used to perform high-dimensional space mapping on the characteristic parameter sets of the plurality of second blood pressure measurement sites to obtain a high-dimensional feature space of each second blood pressure measurement site, wherein the kernel function conversion formula is: , Indicate point and Point The similarity in the high-dimensional space after transformation by the kernel function, express and are two points in the original data space, representing the feature set of two different blood pressure measurement points. express and The Euclidean distance between represents the kernel width parameter, Indicates that it is processed by exponential function and The Euclidean distance between Adjust the sensitivity to obtain a similarity measure after nonlinear mapping;

[0055] (2) Based on the kernel attention mechanism, the attention weight is calculated for the high-dimensional feature space of each second blood pressure measurement site to obtain the attention feature weight of each second blood pressure measurement site;

[0056] (3) Calculate the prior probability density of each second blood pressure measurement site according to the attention feature weight;

[0057] (4) Adjusting the influence weight of the first blood pressure state prediction strategy according to the prior probability density to obtain a second blood pressure state prediction strategy for each second blood pressure measurement site.

[0058] Specifically, the characteristic parameter set of the second blood pressure measurement site is mapped to a high-dimensional space through the kernel function conversion formula. Through nonlinear transformation, the original characteristic space is mapped to a higher-dimensional space to better reveal the internal structure and relationship of the data. The Gaussian kernel function calculates the Euclidean distance between two points x and y in the original data space and converts this distance through an exponential function to obtain a measure that reflects their similarity in the high-dimensional space. The parameter is the kernel width, which determines the degree of dispersion after mapping to the high-dimensional space and has an important impact on the performance of the algorithm. For example, suppose there are two feature sets x and y of blood pressure measurement points. These two points may represent the blood pressure measurement results of the same user at different time points, including parameters such as blood pressure value and pulse wave velocity. Through kernel function transformation, these features are mapped to a higher-dimensional space to capture the complex nonlinear relationship that may exist between these features. The attention weight is calculated for the high-dimensional feature space of each second blood pressure measurement site based on the kernel attention mechanism. The most critical features for predicting blood pressure status are identified in the high-dimensional feature space, and higher attention weights are assigned to these features to improve the accuracy and robustness of the prediction model. For example, if the correlation between pulse wave velocity and abnormal blood pressure is identified as particularly strong in the high-dimensional feature space, then this feature will receive a higher attention weight. According to the attention feature weight, the prior probability density is calculated for each second blood pressure measurement site. The feature importance determined by the attention mechanism is used to estimate the probability distribution of blood pressure status, providing a quantitative basis for subsequent predictions. For example, by analyzing the attention weights, it is found that the blood pressure value at a certain measurement site is more likely to be high under certain conditions. This information will be used to calculate the prior probability density of the measurement site. Based on the prior probability density, the influence weight of the first blood pressure state prediction strategy is adjusted to obtain the second blood pressure state prediction strategy for each second blood pressure measurement site. The original prediction strategy will be adjusted based on the new insights obtained from the high-dimensional feature space to ensure that the prediction results reflect both the latest analysis of the blood pressure measurement data and can adapt to individual differences and changing environmental conditions. For example, if it is found that the prior probability density of blood pressure at a certain measurement site in winter shows a higher risk of abnormality, the corresponding prediction strategy will be adjusted to more sensitively capture blood pressure abnormalities under similar conditions.

[0059] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0060] (1) performing rectangular segmentation on the plurality of second blood pressure measurement sites to obtain a plurality of rectangular segmentation regions;

[0061] (2) Perform global search on multiple rectangular segmentation regions through the rectangular segmentation global optimization algorithm to obtain the minimum prediction error parameter vector for each rectangular segmentation region;

[0062] (3) Based on the minimized prediction error parameter vector, the second blood pressure state prediction strategy of each second blood pressure measurement site is fused to generate a target blood pressure state prediction strategy for the blood pressure measurement system.

[0063] Specifically, a plurality of second blood pressure measurement sites are rectangularly segmented, and the physical space or feature space is divided into a plurality of small regions, so that each region can be analyzed and processed more carefully. For example, if a system including two measurement sites, the wrist and the upper arm, is considered, the rectangular segmentation can divide the measurement data distribution area of ​​each site into a plurality of small rectangular regions based on the blood pressure measurement range or the distribution characteristics of the measurement value of each site. These regions can be divided according to the high and low blood pressure values, the early and late measurement time or other relevant parameters. The rectangular segmentation region is globally searched by the rectangular segmentation global optimization algorithm, and the optimal model parameters are found in each divided small region to minimize the prediction error. The algorithm used in the global search process may include genetic algorithm, particle swarm optimization or other global optimization techniques, aiming to find the parameter vector that minimizes the prediction error. For example, if the blood pressure measurement values ​​are mainly concentrated in the higher range in a certain rectangular segmentation region, the global search algorithm will try to adjust the parameters of the model in the region, such as adjusting the coefficients or weights that affect the blood pressure prediction, so as to ensure that the prediction of the blood pressure state in this specific region is as close as possible to the actual measurement value. According to the minimized prediction error parameter vector obtained for each rectangular segmentation area, the second blood pressure state prediction strategy for each second blood pressure measurement site is fused, and the prediction strategies optimized for different regions are integrated into a unified blood pressure state prediction strategy. The fusion should not only consider the prediction accuracy within each region, but also the coherence between different regions and the consistency of the overall prediction strategy. For example, in different rectangular areas of the upper arm measurement site, each area may obtain different optimized model parameters based on its specific blood pressure distribution characteristics. When fusing these strategies, it is necessary to ensure that the prediction strategy can smoothly transition from one area to another to avoid sudden changes in the prediction results. The target blood pressure state prediction strategy of the blood pressure measurement system finally generated has good adaptability and flexibility while ensuring high accuracy.

[0064] In a specific embodiment, the process of executing step S105 may specifically include the following steps:

[0065] (1) Based on the target blood pressure state prediction strategy, the blood pressure measurement system collects blood pressure data to obtain blood pressure time series data;

[0066] (2) Inputting the blood pressure time series data into a preset initial Volterra functional series model to extract the abnormal fluctuation period and obtain the target abnormal fluctuation period, wherein the initial Volterra functional series model includes: , Indicates at time The blood pressure output is the blood pressure value predicted by the Volterra functional series model based on the input blood pressure time series data. represents the constant term, which represents the baseline value of the blood pressure prediction model, that is, the default blood pressure output when there is no input. It represents the order of the Volterra series, which refers to the depth of considering the influence of past blood pressure changes in the prediction process. The order representative model will take into account the past Blood pressure data at each time point, Indicates The Volterra kernel function of order is used to describe the impact of blood pressure changes at different time points in the past on the current blood pressure prediction value. The parameter represents a specific time delay, Indicates the time delay parameter, representing the time from the current time The length of time to look back into the past, that is, the blood pressure change data after considering the time delay, represents the differential element, which means integrating over all possible time delays to calculate the total effect of past blood pressure changes on the blood pressure prediction at the current moment;

[0067] (3) Perform blood pressure anomaly detection on the target abnormal fluctuation period through the anomaly detection judgment function to obtain the target blood pressure anomaly detection result, wherein the anomaly detection judgment function is: , Indicates at time The anomaly detection result is a binary output, 1 means an anomaly is detected, and 0 means no anomaly is detected. Indicates at time The predicted blood pressure value is calculated by the Volterra functional series model. Indicates at time The reference blood pressure value, Represents the abnormality detection threshold, which is used to determine whether the difference between the predicted blood pressure value and the reference blood pressure value is large enough to be considered abnormal, and is used to distinguish normal blood pressure fluctuations from potential abnormal blood pressure changes.

[0068] Specifically, blood pressure data is collected from the blood pressure measurement system based on the target blood pressure state prediction strategy. For example, for a certain target user, the system may choose to perform more frequent measurements in a specific time period of the day based on past blood pressure measurement records and the patient's health status, in order to capture possible abnormal blood pressure fluctuations. These blood pressure time series data include not only the blood pressure value itself, but also related parameters such as pulse wave velocity and blood flow. The blood pressure time series data is input into the preset initial Volterra functional series model, which processes the blood pressure data through mathematical formulas to extract possible abnormal fluctuation periods. The Volterra model can handle the nonlinear characteristics in time series data and make the prediction more accurate by considering the impact of past blood pressure changes on the current blood pressure prediction value. For example, if the model analysis finds that the blood pressure value suddenly increases or decreases in a certain time period, and this change is significantly different from the past trend, then this time period may be marked as an abnormal fluctuation period. In order to implement anomaly detection specifically, the marked abnormal fluctuation period is further judged using anomaly detection judgment function, which compares the difference between the predicted blood pressure value and the reference blood pressure value and sets a threshold To determine whether there is an abnormality. The threshold needs to be determined based on the patient's specific situation and past blood pressure data to ensure that neither true abnormalities are missed nor normal blood pressure fluctuations are mistakenly judged as abnormalities. For example, for a patient with long-term hypertension, the normal fluctuation range of blood pressure may be different from that of a healthy person, so when performing abnormality detection, the threshold It needs to be adjusted accordingly to suit the patient's specific situation. Finally, the target blood pressure abnormality detection result is obtained.

[0069] In a specific embodiment, the process of executing step S106 may specifically include the following steps:

[0070] (1) Using the adaptive bat algorithm, the model parameters of the initial Volterra functional series model are optimized according to the target blood pressure abnormality detection results to obtain the optimal model parameter solution;

[0071] (2) Update the model parameters of the initial Volterra functional series model by optimizing the model parameter solution to obtain the target Volterra functional series model;

[0072] (3) The blood pressure measurement system is continuously tested for blood pressure anomalies using the target Volterra functional series model, and the continuous blood pressure anomaly detection results are output.

[0073] Specifically, the adaptive bat algorithm is used to optimize the model parameters of the initial Volterra functional series model according to the target blood pressure anomaly detection results. The objective function of the optimization problem is defined according to the target blood pressure anomaly detection results. The error between the model-predicted blood pressure value and the actual measured value is minimized. For example, if the blood pressure data collected by the blood pressure measurement system shows obvious abnormal fluctuations in a certain period of time, these data and their abnormal markers will be used as input to guide the adaptive bat algorithm to optimize the parameters of the initial Volterra functional series model. In this process, the algorithm will evaluate the prediction performance of the model under different parameter settings, continuously update the position of individual bats (i.e., parameter values), and search for the optimal solution in the parameter space by simulating the flight and echolocation behavior of bats. The adaptability of the bat algorithm is reflected in its ability to adjust the search accuracy and step size according to the feedback in the search process, thereby effectively balancing the relationship between exploration (global search) and development (local search) and increasing the probability of finding the global optimal solution. The optimized model parameter solution is applied to the initial Volterra functional series model to update the model parameters. The optimal parameter configuration obtained by the adaptive bat algorithm replaces the corresponding parameters in the initial model to obtain an optimized target Volterra functional series model. The updated model can more accurately reflect the true dynamics of blood pressure data and improve the accuracy and reliability of anomaly detection. Through the target Volterra functional series model, the blood pressure measurement system will perform continuous blood pressure anomaly detection and output the test results. The system will use the updated model to continuously analyze the collected blood pressure time series data and identify and mark abnormal fluctuations in real time.

[0074] The above describes the blood pressure continuous measurement prediction method in the embodiment of the present application. The following describes the blood pressure continuous measurement prediction device in the embodiment of the present application. Figure 2 In the embodiment of the present application, an embodiment of the blood pressure continuous measurement prediction device includes:

[0075] The monitoring module 201 is used to monitor the blood pressure of the target user according to the multiple first blood pressure measurement sites respectively through the blood pressure measurement system, obtain the associated parameter set of each first blood pressure measurement site, and perform spatiotemporal associated node discrimination and node screening on the multiple first blood pressure measurement sites according to the associated parameter set through the K nearest neighbor algorithm to obtain multiple second blood pressure measurement sites;

[0076] A prediction module 202 is used to construct a variable structure dynamic Bayesian network model according to the multiple second blood pressure measurement sites, and predict the blood pressure states of the multiple second blood pressure measurement sites through the variable structure dynamic Bayesian network model to obtain a first blood pressure state prediction strategy for each second blood pressure measurement site;

[0077] An adjustment module 203 is used to perform node prior probability density estimation on multiple second blood pressure measurement sites based on the core attention mechanism to obtain the prior probability density of each second blood pressure measurement site, and adjust the influence weight of the first blood pressure state prediction strategy according to the prior probability density to obtain the second blood pressure state prediction strategy for each second blood pressure measurement site;

[0078] A fusion module 204 is used to fuse the second blood pressure state prediction strategy of each second blood pressure measurement site based on a rectangular segmentation global optimization algorithm to generate a target blood pressure state prediction strategy for the blood pressure measurement system;

[0079] The detection module 205 is used to collect blood pressure data from the blood pressure measurement system based on the target blood pressure state prediction strategy to obtain blood pressure time series data, and input the blood pressure time series data into a preset initial Volterra functional series model to perform blood pressure abnormality detection to obtain a target blood pressure abnormality detection result;

[0080] The output module 206 is used to optimize the initial Volterra functional series model according to the target blood pressure abnormality detection result through the adaptive bat algorithm to obtain the target Volterra functional series model, and perform continuous blood pressure abnormality detection on the blood pressure measurement system through the target Volterra functional series model to output the continuous blood pressure abnormality detection result.

[0081] Through the collaboration of the above components, by continuously monitoring multiple first blood pressure measurement sites and obtaining a set of associated parameters, the target user's blood pressure can be tracked in real time and continuously. This not only helps to capture the dynamic changes of blood pressure in a short period of time, but also can detect abnormal fluctuations in blood pressure in a timely manner. By constructing a variable structure dynamic Bayesian network model and combining the kernel attention mechanism to estimate the node prior probability density, the blood pressure state of each second blood pressure measurement site can be predicted more accurately. It can effectively handle the nonlinear characteristics and time dependence of blood pressure data and improve the accuracy of blood pressure state prediction. Through the spatiotemporal correlation node discrimination and node screening based on the K nearest neighbor algorithm, and the blood pressure state prediction strategy fusion based on the rectangular segmentation global optimization algorithm, it can fully consider the influence of individual differences and environmental factors, and realize the personalization and adaptability of blood pressure prediction. By inputting the blood pressure time series data into the preset initial Volterra functional series model for blood pressure anomaly detection, and using the adaptive bat algorithm for model optimization, it can not only effectively identify abnormal fluctuations in blood pressure, but also adaptively optimize the anomaly detection model to adapt to the complexity of blood pressure changes. It realizes real-time monitoring, accurate prediction and effective abnormality detection of blood pressure, providing users with a comprehensive and efficient blood pressure health management solution. This not only helps users understand their blood pressure status in a timely manner, but also provides data support, thereby improving the accuracy of continuous blood pressure measurement prediction.

[0082] The present application also provides a blood pressure continuous measurement prediction device, which includes a memory and a processor, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the processor executes the steps of the blood pressure continuous measurement prediction method in the above-mentioned embodiments.

[0083] The present application also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of the blood pressure continuous measurement prediction method.

[0084] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0085] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program codes.

[0086] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A blood pressure continuous measurement prediction method, characterized in that: The blood pressure continuous measurement prediction method comprises: The method comprises: performing blood pressure monitoring on a target user according to a plurality of first blood pressure measurement sites respectively through a blood pressure measurement system, obtaining a set of associated parameters for each first blood pressure measurement site, and performing spatiotemporal associated node discrimination and node screening on the plurality of first blood pressure measurement sites according to the associated parameter set through a K nearest neighbor algorithm to obtain a plurality of second blood pressure measurement sites; specifically comprising: initializing the blood pressure measurement site of the target user to obtain a plurality of first blood pressure measurement sites; performing blood pressure monitoring on the target user according to the plurality of first blood pressure measurement sites respectively through a blood pressure measurement system, and obtaining a set of associated parameters for each first blood pressure measurement site respectively, wherein the set of associated parameters comprises: blood pressure value, pulse wave velocity, blood flow, measurement time, user heart rate, user body temperature and environmental parameters; perform parameter contribution rate evaluation and parameter feature sorting on the associated parameter set of each first blood pressure measurement site by using the K nearest neighbor algorithm to obtain a feature parameter set of each first blood pressure measurement site; perform spatiotemporal correlation calculation on the multiple first blood pressure measurement sites according to the feature parameter set by using the dynamic spatiotemporal correlation weight function to obtain a spatiotemporal correlation data set of the multiple first blood pressure measurement sites; perform node screening on the multiple first blood pressure measurement sites according to the spatiotemporal correlation data set by using the complexity-adjusted node screening function to obtain multiple second blood pressure measurement sites; A variable structure dynamic Bayesian network model is constructed according to the multiple second blood pressure measurement sites, and the blood pressure states of the multiple second blood pressure measurement sites are predicted by the variable structure dynamic Bayesian network model to obtain a first blood pressure state prediction strategy for each second blood pressure measurement site; specifically comprising: performing dependency analysis on the multiple second blood pressure measurement sites and constructing a corresponding initial dynamic Bayesian network model; adjusting the node connection weights of the initial dynamic Bayesian network model by a variable structure dynamic weight update function to obtain a variable structure dynamic Bayesian network model; predicting the blood pressure states of the multiple second blood pressure measurement sites by the variable structure dynamic Bayesian network model to obtain a first blood pressure state prediction strategy for each second blood pressure measurement site, wherein the first blood pressure state prediction strategy includes an evaluation of the current blood pressure state and also predicts possible changes in blood pressure trends in the future; Based on the kernel attention mechanism, node prior probability density estimation is performed on the multiple second blood pressure measurement sites respectively to obtain the prior probability density of each second blood pressure measurement site, and the influence weight of the first blood pressure state prediction strategy is adjusted according to the prior probability density to obtain the second blood pressure state prediction strategy for each second blood pressure measurement site; specifically comprising: high-dimensional space mapping of the feature parameter sets of the multiple second blood pressure measurement sites through a kernel function conversion formula to obtain a high-dimensional feature space of each second blood pressure measurement site, wherein the kernel function conversion formula is: , Indicate point and Point The similarity in the high-dimensional space after transformation by the kernel function, express and are two points in the original data space, representing the feature set of two different blood pressure measurement points. express and The Euclidean distance between represents the kernel width parameter, Indicates that it is processed by exponential function and The Euclidean distance between Adjust the sensitivity to obtain a similarity measure after nonlinear mapping; calculate the attention weight of the high-dimensional feature space of each second blood pressure measurement site based on the kernel attention mechanism to obtain the attention feature weight of each second blood pressure measurement site; calculate the prior probability density of each second blood pressure measurement site according to the attention feature weight; adjust the influence weight of the first blood pressure state prediction strategy according to the prior probability density to obtain the second blood pressure state prediction strategy of each second blood pressure measurement site, wherein the second blood pressure state prediction strategy will be adjusted according to the new insights obtained from the high-dimensional feature space to ensure that the prediction result reflects the latest analysis of the blood pressure measurement data and can adapt to individual differences and changing environmental conditions; Based on the rectangular segmentation global optimization algorithm, the second blood pressure state prediction strategy of each second blood pressure measurement site is integrated with the blood pressure state prediction strategy to generate the target blood pressure state prediction strategy of the blood pressure measurement system; specifically comprising: performing rectangular segmentation on the multiple second blood pressure measurement sites to obtain multiple rectangular segmentation areas; performing global search on the multiple rectangular segmentation areas respectively through the rectangular segmentation global optimization algorithm to obtain the minimized prediction error parameter vector of each rectangular segmentation area; based on the minimized prediction error parameter vector, performing blood pressure state prediction strategy integration on the second blood pressure state prediction strategy of each second blood pressure measurement site to generate the target blood pressure state prediction strategy of the blood pressure measurement system; Based on the target blood pressure state prediction strategy, the blood pressure measurement system is subjected to blood pressure data collection to obtain blood pressure time series data, and the blood pressure time series data is input into a preset initial Volterra functional series model to perform blood pressure anomaly detection to obtain a target blood pressure anomaly detection result; specifically comprising: based on the target blood pressure state prediction strategy, the blood pressure measurement system is subjected to blood pressure data collection to obtain blood pressure time series data, wherein the blood pressure time series data includes real-time measurement of blood pressure, real-time recording of measurement environment and user status; the blood pressure time series data is input into a preset initial Volterra functional series model to extract abnormal fluctuation period to obtain a target abnormal fluctuation period, wherein the initial Volterra functional series model includes: , Indicates at time The blood pressure output is the blood pressure value predicted by the Volterra functional series model based on the input blood pressure time series data. represents the constant term, which represents the baseline value of the blood pressure prediction model, that is, the default blood pressure output when there is no input. It represents the order of the Volterra series, which refers to the depth of considering the influence of past blood pressure changes in the prediction process. The order representative model will take into account the past Blood pressure data at each time point, Indicates The Volterra kernel function of order is used to describe the impact of blood pressure changes at different time points in the past on the current blood pressure prediction value. The parameter represents a specific time delay, Indicates the time delay parameter, representing the time from the current time The length of time to look back into the past, that is, the blood pressure change data after considering the time delay, represents a differential element, which represents the integration of all possible time delays to calculate the total impact of past blood pressure changes on the current blood pressure prediction; the blood pressure anomaly detection is performed on the target abnormal fluctuation period through the abnormal detection judgment function to obtain the target blood pressure anomaly detection result, wherein the abnormal detection judgment function is: , Indicates at time The anomaly detection result is a binary output, 1 means an anomaly is detected, and 0 means no anomaly is detected. Indicates at time The predicted blood pressure value is calculated by the Volterra functional series model. Indicates at time The reference blood pressure value, represents the abnormality detection threshold, which is used to determine whether the difference between the predicted blood pressure value and the reference blood pressure value is large enough to be considered abnormal, and is used to distinguish normal blood pressure fluctuations from potential abnormal blood pressure changes; The initial Volterra functional series model is optimized according to the target blood pressure abnormality detection result by using an adaptive bat algorithm to obtain a target Volterra functional series model, and the blood pressure measurement system is continuously tested for blood pressure abnormality by using the target Volterra functional series model to output the continuous blood pressure abnormality detection result.

2. The blood pressure continuous measurement prediction method according to claim 1, characterized in that: The method optimizes the initial Volterra functional series model according to the target blood pressure abnormality detection result by using the adaptive bat algorithm to obtain a target Volterra functional series model, performs continuous blood pressure abnormality detection on the blood pressure measurement system by using the target Volterra functional series model, and outputs the continuous blood pressure abnormality detection result, including: By using an adaptive bat algorithm, the model parameters of the initial Volterra functional series model are optimized according to the target blood pressure abnormality detection result to obtain an optimized model parameter solution; The model parameters of the initial Volterra functional series model are updated by using the optimized model parameter solution to obtain a target Volterra functional series model; The target Volterra functional series model is used to perform continuous blood pressure abnormality detection on the blood pressure measurement system, and a continuous blood pressure abnormality detection result is output.

3. A blood pressure continuous measurement prediction device, characterized in that: The blood pressure continuous measurement prediction device comprises: A monitoring module is used to monitor the blood pressure of a target user according to a plurality of first blood pressure measurement sites through a blood pressure measurement system, obtain a set of associated parameters of each first blood pressure measurement site, and perform spatiotemporal associated node discrimination and node screening on the plurality of first blood pressure measurement sites according to the associated parameter set through a K nearest neighbor algorithm to obtain a plurality of second blood pressure measurement sites; specifically comprising: initializing the blood pressure measurement site of the target user to obtain a plurality of first blood pressure measurement sites; monitoring the blood pressure of the target user according to the plurality of first blood pressure measurement sites through a blood pressure measurement system, and obtaining a set of associated parameters of each first blood pressure measurement site, wherein the set of associated parameters includes : blood pressure value, pulse wave velocity, blood flow, measurement time, user heart rate, user body temperature and environmental parameters; perform parameter contribution rate evaluation and parameter feature sorting on the associated parameter set of each first blood pressure measurement site through the K nearest neighbor algorithm to obtain the characteristic parameter set of each first blood pressure measurement site; perform spatiotemporal correlation calculation on the multiple first blood pressure measurement sites according to the characteristic parameter set through the dynamic spatiotemporal correlation weight function to obtain the spatiotemporal correlation data set of the multiple first blood pressure measurement sites; perform node screening on the multiple first blood pressure measurement sites according to the spatiotemporal correlation data set through the node screening function with complexity adjustment to obtain multiple second blood pressure measurement sites; A prediction module is used to construct a variable structure dynamic Bayesian network model according to the multiple second blood pressure measurement sites, and predict the blood pressure status of the multiple second blood pressure measurement sites through the variable structure dynamic Bayesian network model to obtain a first blood pressure status prediction strategy for each second blood pressure measurement site; specifically comprising: performing dependency analysis on the multiple second blood pressure measurement sites and constructing a corresponding initial dynamic Bayesian network model; adjusting the node connection weights of the initial dynamic Bayesian network model through a variable structure dynamic weight update function to obtain a variable structure dynamic Bayesian network model; predicting the blood pressure status of the multiple second blood pressure measurement sites through the variable structure dynamic Bayesian network model to obtain a first blood pressure status prediction strategy for each second blood pressure measurement site, wherein the first blood pressure status prediction strategy includes an evaluation of the current blood pressure status and also predicts the possible change trend of blood pressure in the future period of time; The adjustment module is used to perform node prior probability density estimation on the multiple second blood pressure measurement sites based on the kernel attention mechanism, obtain the prior probability density of each second blood pressure measurement site, and adjust the influence weight of the first blood pressure state prediction strategy according to the prior probability density to obtain the second blood pressure state prediction strategy for each second blood pressure measurement site; specifically comprising: performing high-dimensional space mapping on the characteristic parameter sets of the multiple second blood pressure measurement sites through a kernel function conversion formula to obtain a high-dimensional feature space of each second blood pressure measurement site, wherein the kernel function conversion formula is: , Indicate point and Point The similarity in the high-dimensional space after transformation by the kernel function, express and are two points in the original data space, representing the feature set of two different blood pressure measurement points. express and The Euclidean distance between represents the kernel width parameter, Indicates that it is processed by exponential function and The Euclidean distance between Adjust the sensitivity to obtain a similarity measure after nonlinear mapping; calculate the attention weight of the high-dimensional feature space of each second blood pressure measurement site based on the kernel attention mechanism to obtain the attention feature weight of each second blood pressure measurement site; calculate the prior probability density of each second blood pressure measurement site according to the attention feature weight; adjust the influence weight of the first blood pressure state prediction strategy according to the prior probability density to obtain the second blood pressure state prediction strategy of each second blood pressure measurement site, wherein the second blood pressure state prediction strategy will be adjusted according to the new insights obtained from the high-dimensional feature space to ensure that the prediction result reflects the latest analysis of the blood pressure measurement data and can adapt to individual differences and changing environmental conditions; A fusion module is used to perform blood pressure state prediction strategy fusion on the second blood pressure state prediction strategy of each second blood pressure measurement site based on a rectangular segmentation global optimization algorithm to generate a target blood pressure state prediction strategy for the blood pressure measurement system; specifically comprising: performing rectangular segmentation on the multiple second blood pressure measurement sites to obtain multiple rectangular segmentation areas; performing global search on the multiple rectangular segmentation areas respectively through a rectangular segmentation global optimization algorithm to obtain a minimized prediction error parameter vector for each rectangular segmentation area; performing blood pressure state prediction strategy fusion on the second blood pressure state prediction strategy of each second blood pressure measurement site based on the minimized prediction error parameter vector to generate a target blood pressure state prediction strategy for the blood pressure measurement system; A detection module, for collecting blood pressure data from the blood pressure measurement system based on the target blood pressure state prediction strategy to obtain blood pressure time series data, and inputting the blood pressure time series data into a preset initial Volterra functional series model to perform blood pressure abnormality detection to obtain a target blood pressure abnormality detection result; specifically comprising: collecting blood pressure data from the blood pressure measurement system based on the target blood pressure state prediction strategy to obtain blood pressure time series data, wherein the blood pressure time series data includes real-time measurement of blood pressure, real-time recording of measurement environment and user status; inputting the blood pressure time series data into a preset initial Volterra functional series model to extract abnormal fluctuation period to obtain a target abnormal fluctuation period, wherein the initial Volterra functional series model includes: , Indicates at time The blood pressure output is the blood pressure value predicted by the Volterra functional series model based on the input blood pressure time series data. represents the constant term, which represents the baseline value of the blood pressure prediction model, that is, the default blood pressure output when there is no input. It represents the order of the Volterra series, which refers to the depth of considering the influence of past blood pressure changes in the prediction process. The order representative model will take into account the past Blood pressure data at each time point, Indicates The Volterra kernel function of order is used to describe the impact of blood pressure changes at different time points in the past on the current blood pressure prediction value. The parameter represents a specific time delay, Indicates the time delay parameter, representing the time from the current time The length of time to look back into the past, that is, the blood pressure change data after considering the time delay, represents a differential element, which represents the integration of all possible time delays to calculate the total impact of past blood pressure changes on the current blood pressure prediction; the blood pressure anomaly detection is performed on the target abnormal fluctuation period through the abnormal detection judgment function to obtain the target blood pressure anomaly detection result, wherein the abnormal detection judgment function is: , Indicates at time The anomaly detection result is a binary output, 1 means an anomaly is detected, and 0 means no anomaly is detected. Indicates at time The predicted blood pressure value is calculated by the Volterra functional series model. Indicates at time The reference blood pressure value, represents the abnormality detection threshold, which is used to determine whether the difference between the predicted blood pressure value and the reference blood pressure value is large enough to be considered abnormal, and is used to distinguish normal blood pressure fluctuations from potential abnormal blood pressure changes; The output module is used to optimize the initial Volterra functional series model according to the target blood pressure abnormality detection result through an adaptive bat algorithm to obtain a target Volterra functional series model, and perform continuous blood pressure abnormality detection on the blood pressure measurement system through the target Volterra functional series model to output the continuous blood pressure abnormality detection result.

4. A blood pressure continuous measurement prediction device, characterized in that: The blood pressure continuous measurement prediction device comprises: a memory and at least one processor, wherein instructions are stored in the memory; The at least one processor calls the instructions in the memory to enable the blood pressure continuous measurement prediction device to execute the blood pressure continuous measurement prediction method according to any one of claims 1-2.

5. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by the processor, the blood pressure continuous measurement prediction method according to any one of claims 1 to 2 is implemented.

Images