A blood pressure risk determination method based on user behavior state and related device

Abstract

The application provides an abnormal segment identification method based on continuous blood pressure information and related equipment, which can more comprehensively capture blood pressure changes at different time scales, effectively improve the detection rate of abnormal intervals, avoid abnormal missed detection caused by single segmentation, and effectively improve the reliability of the identification result. The application comprises: synchronously collecting continuous blood pressure signals; based on a preset blood pressure physiological rhythm prior rule, the continuous blood pressure signals are non-overlapping segmented in three dimensions of long time, medium time and short time to obtain effective segmentation windows; the effective segmentation windows are subjected to abnormal preliminary screening to obtain candidate abnormal intervals; multi-dimensional time sequence features of the continuous blood pressure signals in the candidate abnormal intervals are extracted, and the multi-dimensional time sequence features are spliced to obtain a fusion feature vector.

Description

Technical Field

[0001] This application relates to the field of medical testing technology, and in particular to a method and related equipment for determining blood pressure risk based on user behavior status. Background Technology

[0002] Blood pressure is a key indicator for monitoring cardiovascular health. Its dynamic fluctuations are closely related to daily behavior. Therefore, distinguishing between physiological fluctuations and pathological abnormalities is crucial for blood pressure health management.

[0003] Currently, most blood pressure monitoring technologies rely on single measurements or short-term recordings, making it difficult to capture the dynamic changes in blood pressure over time. They also cannot effectively acquire and correlate user time-series behavioral data with time-series blood pressure data, thus failing to accurately determine whether abnormal fluctuations in a user's blood pressure are caused by pathological or physiological fluctuations, leading to inaccurate blood pressure risk assessment results. Summary of the Invention

[0004] To address the aforementioned technical issues, this application provides a method and related equipment for determining blood pressure risk based on user behavior status.

[0005] The technical solution provided in this application is described below:

[0006] The first aspect of this application provides a method for determining blood pressure risk based on user behavior status, including: Acquire users' time-series behavioral data and time-series blood pressure data; The time-series behavioral data is labeled with the time-effect of the behavioral impact, marking the time of occurrence of the behavior, the period of onset of the effect, and the period of decay of the effect, so as to generate a behavioral time-series feature sequence with time-effect labels; The behavioral temporal feature sequence is time-aligned with the temporal blood pressure data, and the independent impact weight of a single behavior on blood pressure and the coupled superposition impact weight when multiple behaviors occur concurrently are quantified through a temporal causal inference model. A behavioral blood pressure quantification relationship model is constructed based on the independent influence weights and the coupled superimposed influence weights. A personalized dynamic blood pressure baseline for the user is constructed based on normal blood pressure standards, the user's historical blood pressure data, and the aforementioned behavioral blood pressure quantification relationship model. The user's real-time blood pressure data, the personalized dynamic blood pressure baseline, and the behavioral blood pressure quantification relationship model are matched, and physiological fluctuations and pathological abnormalities are distinguished based on the matching results. When a pathological abnormality is identified, the target influence behavior is determined based on the independent influence weights and / or coupled superimposed influence weights. Based on the stated target behavior, intervention recommendations are generated.

[0007] Optionally, the time-series behavioral data is annotated with the time-effect of the behavioral impact, marking the time of occurrence, the period of onset of effect, and the period of attenuation of effect, respectively, to generate a behavioral time-series feature sequence with time-effect labels, including: The time-series behavioral data is subjected to structured preprocessing to obtain standardized time-series behavioral data; The onset delay, duration and decay rate of various behaviors are determined according to the preset behavioral blood pressure time-effect benchmark rules. The occurrence time of each behavior is marked according to the standardized temporal behavior data; The effective period of the behavior is calculated and marked based on the time of occurrence of the behavior and the effective delay. The effect decay period of the behavior is calculated and marked based on the effect onset time, the duration, and the decay rate; Behavioral data labeled with the time of occurrence, the period of onset of effect, and the period of decay of effect are integrated along a unified time axis to generate a temporal feature sequence of behaviors with time-effect labels.

[0008] Optionally, the behavioral temporal feature sequence is time-aligned with the temporal blood pressure data, and the independent impact weight of a single behavior on blood pressure and the coupled superimposed impact weight of multiple concurrent behaviors are quantified using a temporal causal inference model, including: The behavioral temporal feature sequence and the time-series blood pressure data are matched and aligned using a unified time axis with timestamps. The time-aligned behavioral time-series feature sequences and blood pressure time-series data are input into the time-series causal inference model, and the independent influence weights of a single behavior on blood pressure changes are quantified and calculated respectively.

[0009] Determine the temporal overlap relationship of multiple concurrent actions; The temporal overlap relationship of the multi-behavior concurrency is input into the temporal causal inference model to obtain the weight of the coupled superposition effect on blood pressure under the multi-behavior coupling effect.

[0010] Optionally, a personalized dynamic blood pressure baseline is constructed based on normal blood pressure standards, the user's historical blood pressure data, and the behavioral blood pressure quantification model, including: Generate the user's initial static blood pressure baseline based on the user's historical blood pressure data; Based on the aforementioned behavioral blood pressure quantification model, the independent influence weights and coupled superimposed influence weights of behavior on blood pressure are incorporated into the initial static blood pressure baseline to generate a blood pressure correction value that dynamically changes with behavior. A personalized dynamic blood pressure baseline is generated based on the normal blood pressure standard constraints, the user's initial static blood pressure baseline, and the blood pressure correction value.

[0011] Optionally, the user's real-time blood pressure data, the personalized dynamic blood pressure baseline, and the behavioral blood pressure quantification model are matched, and physiological fluctuations are distinguished from pathological abnormalities based on the matching results, including: Obtain the user's real-time blood pressure data at the current moment, and call the corresponding personalized dynamic blood pressure baseline at the current moment to calculate the deviation between the real-time blood pressure and the personalized dynamic blood pressure baseline. Based on the behavioral blood pressure quantification relationship model, the independent influence weights and coupled superimposed influence weights of the user's current time-series behavior and historical time-series behavior, the theoretical fluctuation range of the current blood pressure is predicted. The deviation between the real-time blood pressure and the personalized dynamic blood pressure baseline is compared and matched with the theoretical fluctuation range to obtain the matching result; If the deviation value of the matching result is within the predicted fluctuation range, it is determined to be a physiological fluctuation; if the deviation value exceeds the predicted fluctuation range, it is determined to be a pathological abnormality.

[0012] Optionally, when a pathological abnormality is determined, the target influence behavior is determined based on the independent influence weights and / or coupled superimposed influence weights, including: When blood pressure is determined to be pathologically abnormal, extract all behaviors that occur within the current time window and their corresponding independent influence weights. and / or; Extract the weights of the coupled and superimposed effects of multiple concurrent behaviors within the same time window; The independent influence weights and the coupled superimposed influence weights are sorted, and the behaviors or combinations of behaviors that have the greatest impact on blood pressure abnormalities are selected based on the sorting results. Determine the preset threshold; Identify behaviors or combinations of behaviors with an impact weight greater than a preset threshold, and determine the behaviors or combinations of behaviors as target impact behaviors.

[0013] Optionally, based on the target behavior, intervention suggestion information is output, including: The target time point and target effect intensity of the behavior that causes abnormal effects on blood pressure are determined based on the type of the target behavior, the time of occurrence, the time of onset of effect, and the time of decay of effect. The intervention priority and intensity for the target behavior are calculated based on the behavioral blood pressure quantification relationship model, the independent influence weights, and the coupled superimposed influence weights. Based on the intervention priority and the intervention intensity, a personalized intervention strategy is generated that matches the target behavior. Transform personalized intervention strategies into quantifiable intervention recommendations; The intervention recommendation information will be output.

[0014] A second aspect of this application provides a blood pressure risk assessment device based on user behavior status, the device comprising: The acquisition unit is used to acquire users' time-series behavioral data and time-series blood pressure data; The generation unit is used to annotate the time-series behavioral data with behavioral impact timeliness, marking the time of occurrence of the behavior, the period of onset of effect, and the period of decay of effect, respectively, so as to generate a behavioral time-series feature sequence with timeliness labels; The quantization unit is used to time-align the temporal behavioral feature sequence with the temporal blood pressure data, and quantify the independent influence weight of a single behavior on blood pressure and the coupled superposition influence weight when multiple behaviors occur concurrently through a temporal causal inference model. The first construction unit is used to construct a behavioral blood pressure quantification relationship model based on the independent influence weights and the coupled superimposed influence weights. The second construction unit is used to construct a user-personalized dynamic blood pressure baseline based on normal blood pressure standards, user's historical blood pressure data, and the behavioral blood pressure quantification relationship model. The matching unit is used to match the user's real-time blood pressure data, the personalized dynamic blood pressure baseline, and the behavioral blood pressure quantification relationship model, and to distinguish between physiological fluctuations and pathological abnormalities based on the matching results. The determination unit is used to determine the target influence behavior based on the independent influence weights and / or coupled superimposed influence weights when the abnormality is determined to be pathological. The output unit is used to output intervention suggestion information based on the target influencing behavior.

[0015] A third aspect of this application provides a blood pressure risk assessment device based on user behavior status, the device comprising: Processor, memory, input / output units, and bus; The processor is connected to the memory, the input / output unit, and the bus; The memory stores a program, which the processor invokes to perform the method as described in the first aspect and any one of the first aspects.

[0016] A fourth aspect of this application provides a computer-readable storage medium on which a program is stored, which, when executed on a computer, performs the method as described in the first aspect and any one of the first aspects.

[0017] As can be seen from the above technical solutions, this application has the following beneficial effects: 1. This application obtains user time-series behavior and blood pressure data, marks the behavior data with time-series labels, and generates a time-series feature sequence of behavior with time-series labels. After time alignment, it combines the time-series causal inference model to quantify the independent influence weight of a single behavior and the weight of multiple behaviors coupled and superimposed. In this way, the intrinsic relationship between behavior and blood pressure can be obtained, avoiding the misjudgment problem caused by single factor analysis, thereby improving the accuracy of blood pressure risk assessment.

[0018] 2. This application matches users' real-time blood pressure data, personalized dynamic blood pressure baselines, and behavioral blood pressure quantification models to identify the causes of blood pressure fluctuations and effectively distinguish between physiological fluctuations and pathological abnormalities. This can greatly improve the accuracy of blood pressure risk assessment and reduce misjudgments and omissions.

[0019] 3. In the technical solution of this application, when a pathological abnormality is determined to occur, the target influencing behavior that leads to abnormal blood pressure can be accurately located based on the independent influence weight of behavior and the coupled superimposed influence weight, and targeted intervention suggestions can be output. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a schematic diagram of an embodiment of the abnormal segment identification method based on continuous blood pressure information in this application; Figure 2 This is a schematic diagram of another embodiment of the abnormal segment identification method based on continuous blood pressure information in this application; Figure 3 This is a schematic diagram of another embodiment of the abnormal segment identification method based on continuous blood pressure information in this application; Figure 4 This is a schematic diagram of another embodiment of the abnormal segment identification method based on continuous blood pressure information in this application; Figure 5 This is a schematic diagram of another embodiment of the abnormal segment identification method based on continuous blood pressure information in this application; Figure 6 This is a schematic diagram of another embodiment of the abnormal segment identification method based on continuous blood pressure information in this application; Figure 7 This is a schematic diagram of an embodiment of the abnormal segment identification device based on continuous blood pressure information of this application; Figure 8 This is a schematic diagram of another embodiment of the abnormal segment identification device based on continuous blood pressure information of this application; Figure 9 This is a schematic diagram of another embodiment of the abnormal segment identification device based on continuous blood pressure information in this application. Detailed Implementation

[0022] It should be noted that the blood pressure risk assessment method based on user behavior status disclosed in this embodiment does not have a specific limitation on the executing entity, which can be flexibly selected according to the actual application scenario, as long as the functional logic of each step of this method can be realized.

[0023] Specifically, the executing entity can be a single terminal device, such as a smartphone, smartwatch, blood pressure monitor, tablet computer, or personal computer, which has the ability to collect data, process data, and perform logical operations; it can also be a single server device, including a local server, cloud server, or distributed server, which completes the entire process of blood pressure risk assessment by receiving relevant data transmitted from the terminal; or it can be a collaborative system composed of terminal devices and server devices, in which the terminal devices are responsible for collecting and pre-processing user time-series behavioral data and time-series blood pressure data, while the server devices are responsible for in-depth data analysis, model calculation, and intervention suggestion generation. The two interact with each other through wired or wireless communication to jointly complete the execution of this method.

[0024] Regardless of the execution entity used, as long as it can perform all the steps described in this method, such as "acquiring user time-series data, marking the timeliness of behavioral impacts, time alignment, time-series causal inference, model construction, risk assessment, and output of intervention suggestions," it falls within the scope of the technical solution protected by this invention, and the scope of protection of this invention is not limited by the different execution entities.

[0025] Currently, most blood pressure monitoring technologies rely on single measurements or short-term recordings, making it difficult to capture the dynamic changes in blood pressure over time. They also cannot effectively acquire and correlate user time-series behavioral data with time-series blood pressure data, thus failing to accurately determine whether abnormal fluctuations in a user's blood pressure are caused by pathological or physiological fluctuations, leading to inaccurate blood pressure risk assessment results.

[0026] Based on this, this application provides a method and related equipment for determining blood pressure risk based on user behavior status, which can effectively distinguish between physiological fluctuations and pathological abnormalities, reduce misjudgments and missed judgments, and improve the accuracy of blood pressure risk determination.

[0027] Please see Figure 1 This application discloses a method for identifying abnormal segments based on continuous blood pressure information, the method comprising: 101. Obtain users' time-series behavioral data and time-series blood pressure data; 102. The time-series behavioral data is labeled with the time-effect of the behavioral impact, marking the time of occurrence of the behavior, the period of onset of the effect, and the period of decay of the effect, so as to generate a behavioral time-series feature sequence with time-effect labels; 103. Align the behavioral temporal feature sequence with the temporal blood pressure data in time, and quantify the independent influence weight of a single behavior on blood pressure and the coupled superposition influence weight when multiple behaviors occur concurrently through a temporal causal inference model. 104. Construct a behavioral blood pressure quantification relationship model based on the independent influence weights and the coupled superimposed influence weights; 105. Construct a personalized dynamic blood pressure baseline for the user based on normal blood pressure standards, the user's historical blood pressure data, and the aforementioned behavioral blood pressure quantification relationship model; 106. Match the user's real-time blood pressure data, the personalized dynamic blood pressure baseline, and the behavioral blood pressure quantification relationship model, and distinguish between physiological fluctuations and pathological abnormalities based on the matching results; 107. When a pathological abnormality is determined, the target influence behavior is determined based on the independent influence weights and / or coupled superimposed influence weights. 108. Output intervention suggestions based on the target behavior.

[0028] In this embodiment of the application, the user's time-series behavior data and time-series blood pressure data are first obtained; Next, the temporal behavioral data is labeled with the time-dependent impact of the behavior, marking the time of occurrence, the period of onset of effect, and the period of attenuation of effect to generate a temporal feature sequence of behavior with time-dependent labels. Then, the temporal feature sequence of behavior is aligned with the temporal blood pressure data, and the independent impact weight of a single behavior on blood pressure and the coupled superimposed impact weight when multiple behaviors occur concurrently are quantified through a temporal causal inference model. Then, a behavioral blood pressure quantification relationship model is constructed based on the independent impact weight and the coupled superimposed impact weight. Based on the normal blood pressure standard, the user's historical blood pressure data, and the behavioral blood pressure quantification relationship model, a personalized dynamic blood pressure baseline for the user is constructed. After obtaining the blood pressure baseline, the user's real-time blood pressure data, the personalized dynamic blood pressure baseline, and the behavioral blood pressure quantification relationship model are matched, and physiological fluctuations and pathological abnormalities are distinguished based on the matching results. When a pathological abnormality is determined, the target influencing behavior is determined based on the independent impact weight and / or the coupled superimposed impact weight, and intervention suggestion information is output based on the target influencing behavior.

[0029] In step 101, the user's time-series behavioral data and time-series blood pressure data are acquired. The time-series blood pressure data is collected using wearable blood pressure monitoring devices, such as wrist continuous blood pressure monitors or implantable blood pressure monitoring modules, at a frequency of once per minute for a duration of at least 72 hours to ensure data continuity and integrity. The collected blood pressure data includes systolic blood pressure, diastolic blood pressure, and mean arterial pressure. The timestamp corresponding to each set of blood pressure data is recorded to form a time-series blood pressure data sequence, with timestamps accurate to the second for easy alignment with subsequent behavioral data.

[0030] Time-series behavioral data is acquired through multi-source data collection methods, including smart terminals carried by users, smart home devices, and behavioral information actively recorded by users. The collected behavioral data types cover dietary behavior, exercise behavior, and daily routine behavior, and the timestamp of each behavior is recorded to form a time-series behavioral data sequence, ensuring that the behavioral data and time-series blood pressure data are consistent in time dimension.

[0031] After obtaining complete time-series behavioral data and time-series blood pressure data in step 101, step 102 is executed to annotate the time-series behavioral data with the time-dependent effects of the behavior. This involves marking the time of occurrence, the onset period of the effect, and the attenuation period of the effect, respectively, to generate a time-series feature sequence of behaviors with time-dependent labels. Specifically, for each independent behavior in the time-series behavioral data, the complete time-dependent period of the behavior's impact on blood pressure must be determined using medical knowledge to complete the annotation of the three key time nodes and time periods.

[0032] The moment a behavior occurs is the point in time when the behavior begins to be executed, which directly corresponds to the timestamp of the behavior occurrence in the time-series behavior data.

[0033] The onset time is the period from when the behavior occurs until it has a significant effect on blood pressure. The onset time varies for different types of behaviors. For example, the onset time for eating behavior is 15-30 minutes after the behavior occurs, the onset time for exercise behavior is 5-10 minutes after the behavior occurs, and the onset time for medication behavior is 30-60 minutes after the behavior occurs. This time period is determined by statistically analyzing the average onset time of the effect of this type of behavior on blood pressure, and is also adjusted in combination with individual differences among users.

[0034] The attenuation period is the time after a behavior's effect on blood pressure reaches its peak and gradually weakens until it disappears completely. It is also determined according to the type of behavior. For example, the attenuation period for a single moderate-intensity exercise is 1-2 hours after the exercise ends, the attenuation period for a single medication is 4-6 hours after the medication is taken, and the attenuation period for eating behavior is 2-3 hours after eating ends.

[0035] After completing the labeling of the above three time periods, add a time-effect tag to each behavior. The time-effect tag contains four pieces of information: behavior type, behavior occurrence time, start and end time of effect period, and start and end time of effect decay period. Then, arrange all behaviors with time-effect tags in the order of timestamps to form a time sequence feature sequence of behaviors with time-effect tags.

[0036] After generating the behavioral time-series feature sequence with time-sensitive labels in step 102, step 103 is executed to align the behavioral time-series feature sequence with the time-series blood pressure data in time, and quantify the independent impact weight of a single behavior on blood pressure and the coupled superposition impact weight when multiple behaviors occur concurrently through a time-series causal inference model.

[0037] The time alignment process for the database uses timestamps as a benchmark. It matches the onset and decay periods of each behavior's effect in the time-series feature sequence with time-sensitive labels to the corresponding time intervals of blood pressure data in the time-series blood pressure data. This ensures a one-to-one correspondence between the behavior's time-effect period and the blood pressure changes within that period. For example, if the onset period of a dietary behavior is t1-t2 and the decay period is t2-t3, then the time-sensitive label of this behavior is associated and aligned with all blood pressure data within the t1-t3 time-series blood pressure data, forming a "behavior-blood pressure time-series association pair." After time alignment, an LSTM-based causal inference network is introduced. This network can capture the dynamic associations and causal relationships in the time-series data and quantify the weight of the behavior's impact on blood pressure.

[0038] Among them, the independent impact weight quantification of a single behavior refers to analyzing the change in blood pressure data within the effective period of the behavior through a time-series causal inference model, under the premise of excluding interference from other behaviors, and comparing it with the change in blood pressure data during the same period without the behavior, to calculate the independent impact weight of the behavior on blood pressure. The weight value ranges from 0 to 1. The larger the weight value, the stronger the independent impact of the behavior on blood pressure.

[0039] The quantification of the weighted combined influence of multiple concurrent behaviors refers to analyzing the changes in blood pressure data within the time frame of the combined effects of two or more behaviors when their time-effect periods overlap. This analysis combines the independent influence weights of each behavior with a time-series causal inference model to calculate the weighted combined influence of the combined behaviors. This weight is not a simple sum of the independent influence weights of each behavior, but rather considers the synergistic effect between behaviors. For example, the weighted combined influence of "high-salt diet + sedentary lifestyle" is greater than the independent influence weight of a single high-salt diet. The final output includes the independent influence weight matrix for all individual behaviors and the weighted combined influence matrix for all combinations of concurrent behaviors.

[0040] Based on the independent influence weights and coupled superimposed influence weights obtained in step 103, step 104 is executed to construct a behavioral blood pressure quantitative relationship model according to the independent influence weights and coupled superimposed influence weights. The core of this behavioral blood pressure quantitative relationship model is to establish a quantitative correspondence between a single behavior or a combination of concurrent behaviors and blood pressure changes. The model input is the time-series features of the behavior with time-sensitive labels, and the model output is the predicted blood pressure change value corresponding to that behavior.

[0041] In the specific construction process, the independent influence weights and coupled superimposed influence weights obtained in step 103 are first used as model parameters. Combined with the "behavior-blood pressure time-series correlation pair" obtained in step 103 as training data, the LSTM-based regression model is trained. During training, the model parameters are continuously adjusted to keep the error between the predicted blood pressure change and the actual blood pressure change within a preset threshold until the model converges. Simultaneously, a behavior time-effect decay function is introduced into the model to simulate the decay process of the behavior's influence on blood pressure over time. The parameters of this function are determined based on the decay period marked in step 102 and the influence weights quantified in step 103, ensuring that the model can accurately reflect the degree of influence of behavior on blood pressure at different time stages. The final constructed behavior-blood pressure quantification relationship model can accurately predict the blood pressure changes caused by the input behavior time-series features.

[0042] After constructing the behavioral blood pressure quantification model in step 104, step 105 is executed to construct a personalized dynamic blood pressure baseline for the user based on the normal blood pressure standard, the user's historical blood pressure data, and the behavioral blood pressure quantification model. First, the normal blood pressure standard adopts the prescribed normal blood pressure range (systolic blood pressure 90-139 mmHg, diastolic blood pressure 60-89 mmHg) as the basic reference standard for baseline construction. The user's historical blood pressure data uses time-series blood pressure data collected in step 101 for more than 72 hours, supplemented by the user's routine blood pressure monitoring data from the past 3-6 months. After removing abnormal data, the average value and fluctuation range of the user's historical blood pressure are calculated as the individual reference basis for the personalized baseline.

[0043] Next, combining the behavioral blood pressure quantification model constructed in step 104, the influence of various user behaviors on their blood pressure is analyzed. For different behavior types and combinations, the normal fluctuation range of the user's blood pressure under the influence of each behavior is calculated. For example, the normal blood pressure fluctuation range after eating, after exercise, and during sleep. Based on this, a personalized dynamic blood pressure baseline is constructed for each user. This baseline is not a fixed value, but a blood pressure range that dynamically adjusts with changes in user behavior. That is, for each time point, the normal blood pressure range corresponding to that time point is predicted by combining the behavioral time sequence characteristics before and after that time point through the behavioral blood pressure quantification model, forming a dynamically changing baseline curve. This baseline not only conforms to the normal blood pressure standard, but also fits the individual blood pressure characteristics and behavioral influence patterns of the user, and can distinguish between normal fluctuations and abnormal changes in blood pressure.

[0044] After completing step 105 to construct the user's personalized dynamic blood pressure baseline, step 106 is executed to match the user's real-time blood pressure data, personalized dynamic blood pressure baseline, and behavioral blood pressure quantification relationship model, and to distinguish between physiological fluctuations and pathological abnormalities based on the matching results.

[0045] First, real-time blood pressure data is collected in real time through a wearable blood pressure monitoring device. The collection frequency is consistent with step 101, which is 1 time / minute. At the same time, the real-time collection timestamp and the user's real-time behavior at the time of collection are recorded to form a real-time "blood pressure-behavior" data pair.

[0046] Then, real-time blood pressure data and real-time behavioral data are input into the behavioral blood pressure quantification model to obtain the predicted blood pressure change value corresponding to the real-time behavior. Simultaneously, the normal blood pressure range corresponding to the real-time timestamp is retrieved from the personalized ambulatory blood pressure baseline. The real-time blood pressure data is matched with the normal blood pressure range of the personalized ambulatory blood pressure baseline, and combined with the predicted blood pressure change value output by the behavioral blood pressure quantification model, a dual judgment is made: if the real-time blood pressure data is within the normal blood pressure range of the personalized ambulatory blood pressure baseline, and the deviation from the predicted blood pressure change value is within a preset threshold, then the blood pressure change is determined to be a physiological fluctuation, i.e., a normal blood pressure fluctuation caused by the current behavior, requiring no intervention; if the real-time blood pressure data exceeds the normal blood pressure range of the personalized ambulatory blood pressure baseline, and the deviation from the predicted blood pressure change value exceeds a preset threshold, while excluding the influence of the current behavior and other concurrent behaviors, then the blood pressure change is determined to be a pathological abnormality, i.e., a blood pressure abnormality caused by disease factors, requiring further analysis and intervention.

[0047] When step 106 determines that the blood pressure is pathological, step 107 is executed to determine the target influencing behavior based on the independent influence weights and / or the coupled superimposed influence weights. The target influencing behavior refers to the behavior that has a significant impact on the pathologically abnormal blood pressure or may aggravate the abnormal blood pressure. The determination process combines the independent influence weights and coupled superimposed influence weights quantified in step 103, as well as the temporal characteristics of the behavior corresponding to the blood pressure abnormality segment confirmed in step 106.

[0048] Specifically, the process begins by extracting all behaviors corresponding to the abnormal blood pressure segment. The independent or coupled influence weights of these behaviors are then retrieved, and behaviors with weights greater than a preset threshold are selected as candidate behaviors that may significantly impact blood pressure abnormalities. Subsequently, these candidate behaviors are further analyzed. Using a behavioral-blood pressure quantification model, the contribution of each candidate behavior to blood pressure changes within the abnormal segment is calculated. The contribution is calculated as the product of the behavior's influence weight and the actual change in blood pressure under the effect of that behavior. The 1-2 behaviors with the highest contribution are then selected as target behaviors.

[0049] For example, if two concurrent behaviors exist within the blood pressure abnormality segment: "high-salt diet" (independent influence weight 0.75) and "sedentary lifestyle" (independent influence weight 0.62), and calculations show that the contribution of high-salt diet to blood pressure abnormalities is greater than that of sedentary lifestyle, then high-salt diet is identified as the target influencing behavior. If only a single behavior exists, such as staying up late, and its independent influence weight is higher than a preset threshold, then that behavior is directly identified as the target influencing behavior.

[0050] It should be noted that if there are no behaviors with a weight higher than the preset threshold within the blood pressure abnormality range, it indicates that the pathological abnormality is mainly caused by the disease itself. In this case, the target influencing behavior is set as "no significant influencing behavior", and the subsequent intervention recommendations will focus on the diagnosis and treatment of the disease.

[0051] After identifying the target influencing behavior in step 107, proceed to step 108 to output intervention recommendations based on the target influencing behavior. The output of intervention recommendations must consider the type of the target influencing behavior, the pattern of its impact on blood pressure, and the user's individual characteristics to ensure the recommendations are targeted and actionable. If the target influencing behavior is a single behavior, the intervention recommendations will primarily focus on adjusting that behavior. For example, regarding a high-salt diet, it is recommended to reduce daily salt intake, avoid consuming pickled foods, processed meats, and other high-salt foods, increase the intake of fresh vegetables and fruits, and regularly monitor blood pressure to observe changes in blood pressure after dietary adjustments.

[0052] If the target influencing behavior is a combination of concurrent behaviors, such as "high-salt diet + sedentary lifestyle," then the intervention recommendations should address adjustments to both behaviors. For example, while adjusting the diet, it is recommended to get up and move around for 5-10 minutes every hour of sitting, and to engage in moderate-intensity exercise 3-5 times a week, with each session lasting at least 30 minutes. If the target influencing behavior is a "non-significantly influencing behavior," then the intervention recommendations should focus on the diagnosis and treatment of pathological blood pressure abnormalities. The intervention recommendations can be delivered in the form of text prompts, voice reminders, etc., pushed through the user's smart device to ensure that the user can promptly receive and implement the intervention recommendations, thereby achieving effective management of blood pressure abnormalities.

[0053] Please refer to Figure 2 According to some embodiments of the present invention, in step 102, the time-series behavioral data is labeled with the time-effect of the behavioral impact, marking the time of occurrence of the behavior, the period of onset of the effect, and the period of attenuation of the effect, respectively, to generate a behavioral time-series feature sequence with time-effect labels. Specifically, this may include, but is not limited to, the following: 201. Perform structured preprocessing on the time-series behavioral data to obtain standardized time-series behavioral data; 202. Determine the onset delay, duration, and decay rate of various behaviors based on the preset behavioral blood pressure time-effect benchmark rules; 203. Mark the occurrence time of each behavior according to the standardized temporal behavior data; 204. Calculate and mark the effective period of the behavior based on the time of occurrence of the behavior and the effective delay; 205. Calculate and mark the attenuation period of the behavior based on the onset time, duration, and attenuation rate; 206. Integrate the behavioral data labeled with the time of occurrence, the period of effect onset, and the period of effect decay along a unified time axis to generate a behavioral time-series feature sequence with time-effect labels.

[0054] In this embodiment of the application, the time-series behavioral data is subjected to structured preprocessing to obtain standardized time-series behavioral data. Specifically, the time-series behavioral data obtained in step 101 comes from multiple data collection channels. The data formats and field definitions output from different channels differ; therefore, structured preprocessing is necessary to standardize the data, ultimately outputting standardized time-series behavioral data with a unified structure, standardized format, and complete fields.

[0055] After completing structured preprocessing and obtaining standardized time-series behavioral data, the onset delay, duration, and decay rate of various behaviors are determined according to preset behavioral blood pressure timeliness benchmark rules. These preset behavioral blood pressure timeliness benchmark rules are a standardized rule library built based on medical knowledge and behavioral physiology research results. This rule library defines corresponding timeliness parameter ranges for each type of behavior encoded in step 201, while also reserving an interface for adjusting individual user differences. Parameters can be fine-tuned based on user age, underlying diseases, physical condition, and other characteristics to ensure the accuracy and relevance of the timeliness parameters. Specifically, the onset delay refers to the time required for a behavior to have a significant impact on blood pressure; the onset delay varies significantly among different types of behaviors. The duration refers to the effective duration of the behavior's impact on blood pressure, i.e., the time from the onset of the effect to the beginning of its decay. The decay rate refers to the rate at which the intensity of the effect on blood pressure gradually weakens over time after reaching its peak, expressed as the decrease in intensity per unit time.

[0056] In practice, for each type of behavior in the standardized time-series behavioral data, the corresponding initial values ​​of onset delay, duration and decay rate are matched from the preset behavioral blood pressure time-effect benchmark rule library, and then fine-tuned in combination with individual user characteristics to finally determine the time-effect parameters corresponding to each type of behavior.

[0057] After determining the timeliness parameters of various behaviors, the occurrence time of each behavior is marked based on standardized time-series behavior data. The occurrence time of the behavior is the core benchmark for subsequent calculations of the effective period and the attenuation period of the effect. The marking process is strictly based on the "occurrence timestamp" in the standardized time-series behavior data to ensure the accuracy of the marking.

[0058] Specifically, each behavior record in the standardized time-series behavioral data contains a unique timestamp, corresponding to the specific time when the behavior began. For example, if the timestamp of a dietary behavior record is "202X-XX-XX 12:05:30", then the time when the dietary behavior occurred is marked as the time corresponding to that timestamp. For continuous behaviors, such as sleep or prolonged exercise, the timestamp is the start time of the behavior, not the end time. For example, if the timestamp of the sleep behavior is "202X-XX-XX 22:30:00", even if the sleep behavior lasts for 8 hours, its occurrence time is still marked as 22:30:00. During the labeling process, a "behavior occurrence time" tag is added to each standardized behavioral data record. The tag format is consistent with the timestamp, and it is also associated with the behavior's code, behavior type, and other information to ensure that each behavior record corresponds to a unique behavior occurrence time tag.

[0059] After marking the time of occurrence of the behavior, the onset period of the behavior is calculated and marked based on the time of occurrence and the onset delay. The onset period refers to the time during which the behavior begins to have a significant effect on blood pressure. Its calculation is based on the time of occurrence and the onset delay of the behavior, while also taking into account the characteristics of the behavior type.

[0060] The specific calculation method is as follows: taking the time when the behavior occurs as the starting point, adding the corresponding onset delay of the behavior, we get the onset time, which is the time when the behavior begins to have a significant effect on blood pressure. The end point of the onset period is the time when the effect of the behavior on blood pressure reaches its peak. This peak time is calculated by the onset delay and duration (onset time + duration / 2), that is, the onset period is [onset time, onset time + duration / 2].

[0061] For example, if a certain exercise activity occurs at 14:00:00, with a corresponding onset delay of 8 minutes and a duration of 40 minutes, then the onset time of this activity is 14:08:00, and the peak time is 14:08:00 + 20 minutes = 14:28:00. Therefore, the effective period of this exercise activity is marked as 14:08:00-14:28:00. During the marking process, a "effective period" tag is added to each behavior record, clearly indicating the start and end times of the period, and simultaneously linking it to the onset delay parameter of the behavior. This facilitates subsequent verification and adjustment, ensuring that the marked effective period corresponds to the actual impact of the behavior on blood pressure.

[0062] Next, the attenuation period of the behavior's effect is calculated and marked based on the onset time, duration, and attenuation rate. The attenuation period refers to the time during which the effect of the behavior on blood pressure gradually weakens until it disappears completely after reaching its peak. Its calculation needs to combine the end of the onset period, the remaining duration, and the attenuation rate to ensure an accurate reflection of the decline process of the behavior's effect.

[0063] The specific calculation method is as follows: take the end of the effective period (peak time) as the starting point, and calculate the time point when the effect completely disappears in combination with the decay rate, that is, the decay end time = peak time + (1 / decay rate). Therefore, the effect decay period is [peak time, decay end time].

[0064] It should be noted that if the decay endpoint time exceeds the actual duration of the behavior (e.g., the behavior itself has ended but the effect has not yet subsided), the decay endpoint time will be adjusted based on the behavior's end time, combined with the decay rate, to ensure that the decay period matches the actual occurrence scenario of the behavior. For example, if the peak time of an dietary behavior's effect is 12:45:00, with 30 minutes remaining and a decay rate of 0.015 / minute, then 1 / decay rate ≈ 66.7 minutes. Therefore, the decay endpoint time is 12:45:00 + 66.7 minutes ≈ 13:51:42, and the decay period is marked as 12:45:00-13:51:42. During the marking process, a "decay period" tag is added to each behavior record, indicating the start and end times of the period, and associating it with parameters such as the decay rate and peak time to form complete time-sensitive labeling information.

[0065] Finally, the behavioral data, labeled with the time of occurrence, the period of onset of action, and the period of attenuation of action, are integrated along a unified timeline to generate a behavioral time-series feature sequence with time-sensitive labels. The unified timeline uses the timestamps of the time-series blood pressure data from step 101 as a benchmark, ensuring consistent time precision with the blood pressure data and guaranteeing accurate alignment between subsequent behavioral data and blood pressure data.

[0066] The integration process includes three stages: Data Association: Each piece of data with time-sensitive information (i.e., the time of occurrence, the period of effect onset, and the period of effect decay) is associated with corresponding time nodes on a unified timeline according to the chronological order of occurrence. Conflict Handling: When the time-sensitive periods of multiple behaviors overlap, the complete time-sensitive information of each behavior is retained, and a concurrent behavior identifier is added to clearly identify the combination of behaviors within the overlapping time period, providing a basis for subsequent quantification of the weight of the concurrent impact of multiple behaviors. For example, if the onset time of behavior A is 14:08:00-14:28:00, and the onset time of behavior B is 14:15:00-14:35:00, with an overlap of 14:15:00-14:28:00, the integration process retains the complete time-sensitive information of behaviors A and B respectively, and marks them as "14:15:00-14:28:00 behavior A and behavior B concurrent". Feature integration: The core information of each behavioral data point is integrated into a complete feature record, arranged in chronological order along a unified timeline, forming a behavioral time-series feature sequence with time-sensitive labels.

[0067] Please refer to Figure 3 According to some embodiments of the present invention, in step 103, the behavioral temporal feature sequence is time-aligned with the temporal blood pressure data, and the independent influence weight of a single behavior on blood pressure and the coupled superimposed influence weight when multiple behaviors occur concurrently are quantified through a temporal causal inference model. Specifically, this may include, but is not limited to, the following: 301. Match and align the behavioral temporal feature sequence with the temporal blood pressure data according to a unified time axis using timestamp matching; 302. Input the time-aligned behavioral time-series feature sequence and blood pressure time-series data into the time-series causal inference model, and quantify the independent influence weight of a single behavior on blood pressure changes.

[0068] 303. Determine the temporal overlap relationship of multiple concurrent actions; 304. Input the temporal overlap relationship of the multi-behavior concurrency into the temporal causal inference model to obtain the weight of the coupled superposition effect of blood pressure under the multi-behavior coupling effect.

[0069] In this embodiment, after completing timestamp matching and alignment, the time-aligned behavioral temporal feature sequence and blood pressure temporal data are input into the temporal causal inference model to quantify and calculate the independent influence weight of a single behavior on blood pressure changes. The independent influence weight of a single behavior refers to the degree of influence of the behavior on the user's blood pressure changes when the behavior acts alone on the user, excluding interference from other behaviors. The weight value ranges from 0 to 1. The larger the weight value, the stronger the independent influence of the behavior on blood pressure changes, and vice versa.

[0070] Specifically, firstly, the single-behavior screening criteria are determined. From the time-aligned combination set output in step 301, "behavior timeliness-blood pressure" association pairs that only have the timeliness impact of a single behavior and have no concurrent behaviors are selected as training and calculation data for quantifying the independent impact weight of a single behavior. For example, there are association pairs that only have dietary behavior, no exercise, medication, or other concurrent behaviors, and association pairs that only have a single exercise behavior. Next, a time-series causal inference model is constructed. This application adopts a causal inference network based on Long Short-Term Memory (LSTM), which can accurately capture the dynamic associations and causal relationships of time-series data and adapt to the time-series characteristics of behavior timeliness and blood pressure changes.

[0071] The model input consists of time-series features of a single behavior and time-series blood pressure data. The model output is the independent impact weight of that behavior on blood pressure changes. Subsequently, the selected single-behavior "behavior duration-blood pressure" association pairs are divided into training and testing sets and input into the time-series causal inference model for training. During training, the error between the actual blood pressure change value and the blood pressure change value predicted by the model during the behavior's period is used as the loss function. The model parameters are continuously adjusted using the gradient descent method until the model converges. Finally, the test set data is input into the trained time-series causal inference model to verify the model's quantitative accuracy. For models that pass the verification, all single-behavior "behavior duration-blood pressure" association pairs are input, and the independent impact weight of each type of single behavior and each single behavior record on blood pressure changes is calculated to form a single-behavior independent impact weight matrix. The matrix clearly marks the behavior code, behavior type, corresponding blood pressure index, and corresponding independent impact weight value.

[0072] After quantifying the independent impact weights of individual behaviors, the temporal overlap relationships of concurrent behaviors are further determined. Concurrent behaviors refer to the overlapping time periods of two or more behaviors, meaning that multiple behaviors simultaneously affect a user's blood pressure within the same time interval. The temporal overlap relationships of concurrent behaviors are then input into a temporal causal inference model to obtain the weight of the coupled and superimposed impact on blood pressure under the combined effect of multiple behaviors. The weight of the coupled and superimposed impact of multiple behaviors refers to the comprehensive degree of influence on changes in a user's blood pressure after the interaction of various behaviors during concurrent behavior. This weight is not a simple sum of the independent impact weights of each behavior, but rather a comprehensive calculation combining the coupling effect between behaviors, the temporal overlap relationship, and the time parameters.

[0073] Specifically, first, the set of concurrent temporal overlap relationships of multiple behaviors output in step 303 is retrieved, and the temporal overlap relationship report of each group of concurrent behavior combinations is extracted. Simultaneously, the independent influence weight of each behavior in the combination calculated in step 302 is retrieved. This information is integrated into a multi-behavior concurrent feature vector, which serves as the input data for the temporal causal inference model. Next, using the temporal causal inference model trained in step 302, the model's input layer is fine-tuned by adding three new input dimensions: the number of concurrent behaviors, the duration of temporal overlap, and the type of behavior coupling, to adapt to the feature input requirements of multi-behavior concurrency. The model output is the coupled superimposed influence weight of the concurrent behavior combination on blood pressure changes. Then, all multi-behavior concurrent feature vectors are input into the fine-tuned temporal causal inference model to calculate the coupled superimposed influence weight of each group of concurrent behavior combinations. During the calculation, the length of the temporal overlap period, the timeliness of each behavior within the overlap period, and the coupling effect between behaviors are considered.

[0074] Finally, the calculated coupling superposition influence weights are verified. The verification method is to compare the actual blood pressure change values ​​of concurrent behavior combinations with the blood pressure change values ​​predicted by the model based on the coupling superposition influence weights to ensure that the error is controlled within the preset threshold. After the verification is passed, a multi-behavior concurrent coupling superposition influence weight matrix is ​​formed. The matrix is ​​marked with concurrency identifier, concurrent behavior combination, temporal overlap relationship, independent influence weight of each behavior and coupling superposition influence weight, and the complete set of multi-behavior coupling superposition influence weights is output.

[0075] Please refer to Figure 4 According to some embodiments of the present invention, in step 105, a personalized dynamic blood pressure baseline for the user is constructed based on normal blood pressure standards, the user's historical blood pressure data, and the behavioral blood pressure quantification relationship model. This may specifically include, but is not limited to, the following: 401. Generate the user's initial static blood pressure baseline based on the user's historical blood pressure data; 402. Based on the aforementioned behavioral blood pressure quantification relationship model, the independent influence weights and coupled superimposed influence weights of behavior on blood pressure are incorporated into the initial static blood pressure baseline to generate a blood pressure correction value that dynamically changes with behavior. 403. Generate a personalized dynamic blood pressure baseline for the user based on the normal blood pressure standard constraints, the user's initial static blood pressure baseline, and the blood pressure correction value.

[0076] In this embodiment, an initial static blood pressure baseline is generated based on the user's historical blood pressure data. The initial static blood pressure baseline is the foundation for the personalized dynamic blood pressure baseline. It is a static reference range that reflects the user's baseline blood pressure level, constructed based on statistical analysis of the user's historical blood pressure data. It does not consider the dynamic impact of behavior on blood pressure and only reflects the inherent characteristics of the user's individual blood pressure.

[0077] Specifically, firstly, the user's historical blood pressure data, obtained and preprocessed in step 101, is retrieved. This data includes at least 72 hours of continuous time-series blood pressure data, supplemented with the user's routine blood pressure monitoring data from the past 3-6 months to ensure data completeness and representativeness. Then, the retrieved historical blood pressure data undergoes further screening and purification, removing invalid data and extreme outliers. Linear interpolation is used to supplement missing data, ensuring the continuity of the data sequence.

[0078] Furthermore, statistical analysis is performed on the purified historical blood pressure data to calculate core statistical indicators such as systolic blood pressure, diastolic blood pressure, and mean arterial pressure. Finally, an initial resting blood pressure baseline report for the user is generated. It should be noted that this initial resting blood pressure baseline only reflects the user's baseline blood pressure status without specific behavioral interventions.

[0079] After generating the user's initial static blood pressure baseline, based on the behavioral blood pressure quantification relationship model, the independent influence weight and coupled superimposed influence weight of behavior on blood pressure are incorporated into the initial static blood pressure baseline to generate a blood pressure correction value that dynamically changes with behavior.

[0080] The blood pressure correction value is the core of achieving dynamic baseline adjustment. Its core logic involves using a behavioral blood pressure quantification model to transform the behavioral impact weights quantified in steps 302 and 304 into specific blood pressure changes under the corresponding behaviors. This value is then used to correct the initial static blood pressure baseline, ensuring the baseline adapts to the dynamic impact of behavior on blood pressure. After generating the blood pressure correction value, a personalized dynamic blood pressure baseline is generated based on normal blood pressure standards, the user's initial static blood pressure baseline, and the blood pressure correction value.

[0081] The purpose of a personalized dynamic blood pressure baseline is to "use an initial static baseline as a basis, dynamically adjust based on blood pressure correction values, and use normal blood pressure standards as boundary constraints" to achieve dynamic adjustment of the baseline as user behavior changes, thus conforming to the individual blood pressure characteristics of the user and effectively distinguishing between physiological fluctuations and pathological abnormalities in blood pressure.

[0082] Please refer to Figure 5 According to some embodiments of the present invention, in step 106, the user's real-time blood pressure data, the personalized dynamic blood pressure baseline, and the behavioral blood pressure quantification relationship model are matched, and physiological fluctuations and pathological abnormalities are distinguished based on the matching results. Specifically, this may include, but is not limited to, the following: 501. Obtain the user's real-time blood pressure data at the current moment, and call the corresponding personalized dynamic blood pressure baseline at the current moment to calculate the deviation between the real-time blood pressure and the personalized dynamic blood pressure baseline. 502. Based on the aforementioned behavioral blood pressure quantification relationship model, the independent influence weights and coupled superimposed influence weights of the user's current time-series behavior and historical time-series behavior, predict the theoretical fluctuation range of the current blood pressure; 503. Compare and match the deviation between the real-time blood pressure and the personalized dynamic blood pressure baseline with the theoretical fluctuation range to obtain the matching result; 504. If the deviation value of the matching result is within the predicted fluctuation range, it is determined to be a physiological fluctuation; if the deviation value exceeds the predicted fluctuation range, it is determined to be a pathological abnormality.

[0083] In this embodiment of the application, the user's real-time blood pressure data at the current moment is obtained, and the corresponding personalized dynamic blood pressure baseline at the current moment is called to calculate the deviation value between the real-time blood pressure and the personalized dynamic blood pressure baseline.

[0084] Specifically, firstly, the system collects the user's current blood pressure data in real time, maintaining the same collection frequency as described above. After collection, the real-time blood pressure data undergoes immediate preprocessing to remove invalid data caused by temporary equipment malfunctions or measurement posture deviations. Then, based on the current collection timestamp, the system retrieves the user's personalized dynamic blood pressure baseline generated in step 403 above and matches it with the dynamic baseline range corresponding to the current moment. If there is no action at the current moment, the initial static blood pressure baseline range is used; if there is a single action or multiple actions occurring concurrently at the current moment, the dynamically adjusted baseline range corresponding to that action is used. Finally, the deviation between the real-time blood pressure and the personalized dynamic blood pressure baseline is calculated. The deviation is calculated based on the baseline median, with systolic blood pressure, diastolic blood pressure, and mean arterial pressure calculated independently. The specific calculation formula is "deviation = |real-time blood pressure value - baseline median value|". Simultaneously, the blood pressure index type, current moment, baseline range, and real-time blood pressure value corresponding to the deviation are recorded, forming a "real-time blood pressure - baseline deviation" correlation data.

[0085] After calculating the deviation value, the theoretical fluctuation range of the current blood pressure is predicted based on the behavioral blood pressure quantification relationship model, the independent influence weights and coupled superimposed influence weights of the user's current time-series behavior and historical time-series behavior.

[0086] Among them, the theoretical fluctuation range is the core reference for determining whether blood pressure fluctuations are normal. Its core logic is to combine the influence weights of current and historical behaviors and predict the normal fluctuation range of blood pressure at the current moment through a behavioral blood pressure quantitative relationship model. In essence, it is a quantitative mapping of the dynamic influence of behavior on blood pressure.

[0087] Specifically, first, the trained and converged behavioral blood pressure quantification relationship model is retrieved, along with the single-behavior independent influence weight matrix output in step 302 and the multi-behavior concurrent coupling superimposed influence weight set output in step 304, ensuring the consistency between model parameters and weight data. Then, the user's current time-series behavior is collected, recording the type, time stage, and duration of the current behavior. Simultaneously, the user's historical time-series behavior is retrieved to clarify the type, time-residual status, and corresponding influence weights of historical behaviors.

[0088] If only a single behavior exists, the independent influence weight and time-effect parameter corresponding to that behavior are retrieved. If multiple behaviors occur concurrently, the coupled superimposed influence weight and time-series overlap parameter corresponding to the concurrent combination are retrieved. If there is no behavior currently but there are residual effects from historical behaviors, the residual influence weight corresponding to the historical behaviors is retrieved. Subsequently, the feature information of the current time-series behavior and the historical time-series behavior, along with their corresponding influence weights, are input into the behavior-blood pressure quantification relationship model. The model, combined with the initial static blood pressure baseline median value from step 401, outputs the theoretical fluctuation range of blood pressure at the current moment. This range is predicted separately for systolic blood pressure, diastolic blood pressure, and mean arterial pressure. Specifically, the calculation is "Theoretical fluctuation range = baseline median value ± (baseline median value × corresponding influence weight × time-effect intensity)," where the time-effect intensity is determined based on the time-effect stage of the current behavior and the residual effects of the behavior, ensuring that the theoretical fluctuation range can reflect the normal blood pressure fluctuation range under the influence of the behavior. Finally, a "behavior-theoretical fluctuation range" association table is generated, clearly marking the current and historical behavior information, corresponding influence weights, time-effect parameters, and the predicted theoretical fluctuation ranges of systolic blood pressure, diastolic blood pressure, and mean arterial pressure.

[0089] After predicting the theoretical fluctuation range, the deviation between real-time blood pressure and the personalized ambulatory blood pressure baseline is compared and matched with the theoretical fluctuation range to obtain the matching result. This step verifies whether the real-time blood pressure deviation is within the normal fluctuation range under behavioral influence through double comparison, providing a clear basis for subsequent fluctuation type determination.

[0090] Specifically, first, the system retrieves the "real-time blood pressure - baseline deviation" correlation data generated in step 501 and the "behavioral - theoretical fluctuation range" correlation table generated in step 502 to ensure that the compared blood pressure indicators correspond one-to-one. Next, using the theoretical fluctuation range as a reference, a dual constraint is applied, combined with the personalized dynamic blood pressure baseline range. Finally, a comparison and matching report is output, clearly indicating the real-time deviation value, theoretical fluctuation range, comparison results, and comparison basis for each blood pressure indicator.

[0091] After obtaining the comparison and matching results, if the deviation value of the matching result is within the predicted fluctuation range, it is determined to be a physiological fluctuation; if the deviation value exceeds the predicted fluctuation range, it is determined to be a pathological abnormality. This step is the core judgment link for identifying abnormal blood pressure segments. It is strictly based on the comparison results mentioned above, combined with the influence of behavior on blood pressure, to distinguish between physiological fluctuations and pathological abnormalities.

[0092] Please refer to Figure 6 According to some embodiments of the present invention, in step 107, when a pathological abnormality is determined, the target influence behavior is determined based on the independent influence weight and / or coupled superimposed influence weight. Specifically, this may include, but is not limited to, the following: 601. When blood pressure is determined to be pathologically abnormal, extract all behaviors that occur within the current time window and their corresponding independent influence weights and / or extract the coupled superimposed influence weights corresponding to multiple behaviors that coexist within the same time window. 602. Sort the independent influence weights and the coupled superimposed influence weights, and select the behaviors or combinations of behaviors that have the greatest impact on blood pressure abnormalities based on the sorting results; 603. Determine the preset threshold; 604. Obtain behaviors or combinations of behaviors with an impact weight greater than a preset threshold, and identify the behaviors or combinations of behaviors as target impact behaviors.

[0093] In this embodiment of the application, when blood pressure is determined to be pathologically abnormal, all behaviors occurring within the current time window and their corresponding independent influence weights are extracted and / or the coupled superimposed influence weights corresponding to multiple behaviors that coexist within the same time window are extracted.

[0094] Specifically, firstly, the definition criteria for the "current time window" are clarified. This time window is centered on the moment when a pathological abnormality is determined in step 504. Combining the continuous characteristics of blood pressure abnormalities and the time-effect pattern of behavior on blood pressure, the time window range is set to "30 minutes before the abnormality determination time to 10 minutes after the abnormality determination time." This range can fully cover the potential behaviors that lead to blood pressure abnormalities while excluding interference from irrelevant historical behaviors, ensuring that the extracted behaviors are directly related to blood pressure abnormalities. Next, based on this time window range, the single behavior independent influence weight matrix output in step 302 and the multi-behavior concurrent coupling superimposed influence weight set output in step 304 are retrieved. At the same time, the time-series behavior data collected by wearable devices and smart terminals are linked to extract all behaviors that occur within this time window. Finally, a "time window-behavior-influence weight" associated dataset is generated, clearly labeling the start and end times of the time window, behavior type, behavior details, and corresponding influence weights, ensuring that each extracted behavior and weight accurately corresponds to the time range of the current pathological blood pressure abnormality.

[0095] After extracting the behaviors and their corresponding weights, the independent influence weights and the coupled influence weights are sorted. Based on the sorting results, the behaviors or combinations of behaviors that have the greatest impact on blood pressure abnormalities are selected. The core of this step is to quantify the degree of influence of different behaviors and combinations of behaviors on pathological blood pressure abnormalities through weight sorting, thereby achieving the screening of core influencing factors. Next, a preset threshold is determined, and behaviors or combinations of behaviors with influence weights greater than the preset threshold are further obtained and identified as target influencing behaviors.

[0096] Please refer to Figure 7 According to some embodiments of the present invention, step 108, which outputs intervention suggestion information based on the target influencing behavior, may specifically include, but is not limited to, the following: 701. Determine the target time point and target effect intensity of the abnormal effect of the behavior on blood pressure based on the type of the target influencing behavior, the time of occurrence, the onset period of effect, and the attenuation period of effect; 702. Calculate the intervention priority and intensity of the target behavior based on the behavioral blood pressure quantification relationship model, the independent influence weight, and the coupled superimposed influence weight; 703. Generate a personalized intervention strategy that matches the target behavior based on the intervention priority and the intervention intensity; 704. Transform personalized intervention strategies into quantifiable intervention recommendations; 705. Output the intervention recommendation information.

[0097] Detailed Explanation of Steps 701-705 In this embodiment, the target time point and target effect intensity of the target influencing behavior are determined based on the type, occurrence time, onset period, and attenuation period of the target influencing behavior. Then, the intervention priority and intensity of the target influencing behavior are calculated based on the behavior-blood pressure quantitative relationship model, independent influence weights, and coupled superimposed influence weights. The core of this step is to clarify the order and intensity of interventions for different target influencing behaviors through quantitative analysis.

[0098] Specifically, the behavioral blood pressure quantification relationship model constructed in step 404, the independent influence weight matrix output in step 302, and the coupled superimposed influence weight set output in step 304 are retrieved first. At the same time, the target effect intensity calculated in step 701 is introduced to construct a two-dimensional calculation system.

[0099] Next, intervention priority is calculated using the target effect intensity as an indicator, combined with a weighted calculation based on the type of behavior influencing the target. The priority score is calculated as: Target Effect Intensity × Behavior Type Weight. The behavior type weight is set according to the reversibility of the behavior's impact on blood pressure; the higher the score, the higher the intervention priority. If multiple target behaviors exist, they are sorted in descending order of priority score, with the highest-priority behavior typically included in the intervention scope first. Then, the intervention intensity is calculated, ranging from 0 to 3, divided into three levels: low (0-1), medium (1-2), and high (2-3). Finally, an intervention priority ranking table and an intervention intensity level table are generated to clarify the priority order of each target behavior.

[0100] After calculating the intervention priority and intensity, a personalized intervention strategy matching the target behavior is generated based on these factors. This personalized intervention strategy is then transformed into quantifiable intervention suggestions. This process converts abstract intervention strategies into concrete, actionable, and quantifiable operational instructions, avoiding vague descriptions and ensuring clear execution by the user. Finally, the intervention suggestions are output for user review.

[0101] Please see Figure 8 The second aspect of this application provides a blood pressure risk assessment device based on user behavior status, the device comprising: Acquisition unit 801 is used to acquire the user's time-series behavior data and time-series blood pressure data; The generation unit 802 is used to annotate the time-series behavior data with the time-effect of the behavior, marking the time when the behavior occurs, the time period during which the effect takes effect, and the time period during which the effect weakens, so as to generate a behavior time-series feature sequence with time-effect labels. The quantization unit 803 is used to time-align the time-series behavioral feature sequence with the time-series blood pressure data, and quantify the independent influence weight of a single behavior on blood pressure and the coupled superposition influence weight when multiple behaviors occur concurrently through a time-series causal inference model. The first construction unit 804 is used to construct a behavioral blood pressure quantification relationship model based on the independent influence weights and the coupled superimposed influence weights. The second construction unit 805 is used to construct a user-personalized dynamic blood pressure baseline based on normal blood pressure standards, user historical blood pressure data and the behavioral blood pressure quantification relationship model. The matching unit 806 is used to match the user's real-time blood pressure data, the personalized dynamic blood pressure baseline, and the behavioral blood pressure quantification relationship model, and to distinguish between physiological fluctuations and pathological abnormalities based on the matching results. The determination unit 807 is used to determine the target influence behavior based on the independent influence weight and / or coupled superimposed influence weight when the abnormality is determined to be pathological. Output unit 808 is used to output intervention suggestion information based on the target influencing behavior.

[0102] Please see Figure 9 This application also provides a blood pressure risk assessment device based on user behavior status, the device comprising: Processor 901, memory 902, input / output unit 903, bus 904; The processor 901 is connected to the memory 902, the input / output unit 903, and the bus 904; The memory 902 stores a program, and the processor 901 calls the program to execute any of the methods described above.

[0103] This application also relates to a computer-readable storage medium on which a program is stored, which, when run on a computer, causes the computer to perform any of the methods described above.

[0104] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0105] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between apparatuses or units, and may be electrical, mechanical, or other forms.

[0106] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0107] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0108] If the integrated unit is implemented as 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 this application, in essence, 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. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A method for determining blood pressure risk based on user behavior status, characterized in that, include: Acquire users' time-series behavioral data and time-series blood pressure data; The time-series behavioral data is labeled with the time-effect of the behavioral impact, marking the time of occurrence of the behavior, the period of onset of the effect, and the period of decay of the effect, so as to generate a behavioral time-series feature sequence with time-effect labels; The behavioral temporal feature sequence is time-aligned with the temporal blood pressure data, and the independent impact weight of a single behavior on blood pressure and the coupled superposition impact weight when multiple behaviors occur concurrently are quantified through a temporal causal inference model. A behavioral blood pressure quantification relationship model is constructed based on the independent influence weights and the coupled superimposed influence weights. A personalized dynamic blood pressure baseline for the user is constructed based on normal blood pressure standards, the user's historical blood pressure data, and the aforementioned behavioral blood pressure quantification relationship model. The user's real-time blood pressure data, the personalized dynamic blood pressure baseline, and the behavioral blood pressure quantification relationship model are matched, and physiological fluctuations and pathological abnormalities are distinguished based on the matching results. When a pathological abnormality is identified, the target influence behavior is determined based on the independent influence weights and / or coupled superimposed influence weights. Based on the stated target behavior, intervention recommendations are generated.

2. The method for determining blood pressure risk based on user behavior status according to claim 1, characterized in that, The time-series behavioral data is labeled with the time-effect of the behavioral impact, marking the time of occurrence, the period of onset of effect, and the period of attenuation of effect, respectively, to generate a behavioral time-series feature sequence with time-effect labels, including: The time-series behavioral data is subjected to structured preprocessing to obtain standardized time-series behavioral data; The onset delay, duration and decay rate of various behaviors are determined according to the preset behavioral blood pressure time-effect benchmark rules. The occurrence time of each behavior is marked according to the standardized temporal behavior data; The effective period of the behavior is calculated and marked based on the time of occurrence of the behavior and the effective delay. The effect decay period of the behavior is calculated and marked based on the effect onset time, the duration, and the decay rate; Behavioral data labeled with the time of occurrence, the period of onset of effect, and the period of decay of effect are integrated along a unified time axis to generate a temporal feature sequence of behaviors with time-effect labels.

3. The method for determining blood pressure risk based on user behavior status according to claim 1, characterized in that, The behavioral time-series feature sequence is time-aligned with the time-series blood pressure data, and the independent impact weight of a single behavior on blood pressure, as well as the coupled and superimposed impact weight of multiple concurrent behaviors, are quantified using a time-series causal inference model, including: The behavioral temporal feature sequence and the time-series blood pressure data are matched and aligned using a unified time axis with timestamps. The time-aligned behavioral time-series feature sequences and blood pressure time-series data are input into the time-series causal inference model, and the independent influence weights of a single behavior on blood pressure changes are quantified and calculated respectively. Determine the temporal overlap relationship of multiple concurrent actions; The temporal overlap relationship of the multi-behavior concurrency is input into the temporal causal inference model to obtain the weight of the coupled superposition effect on blood pressure under the multi-behavior coupling effect.

4. The method for determining blood pressure risk based on user behavior status according to claim 3, characterized in that, A personalized dynamic blood pressure baseline is constructed based on normal blood pressure standards, the user's historical blood pressure data, and the aforementioned behavioral blood pressure quantification model, including: Generate the user's initial static blood pressure baseline based on the user's historical blood pressure data; Based on the aforementioned behavioral blood pressure quantification model, the independent influence weights and coupled superimposed influence weights of behavior on blood pressure are incorporated into the initial static blood pressure baseline to generate a blood pressure correction value that dynamically changes with behavior. A personalized dynamic blood pressure baseline is generated based on the normal blood pressure standard constraints, the user's initial static blood pressure baseline, and the blood pressure correction value.

5. The method for determining blood pressure risk based on user behavior status according to claim 1, characterized in that, The system matches the user's real-time blood pressure data, the personalized dynamic blood pressure baseline, and the behavioral blood pressure quantification model, and distinguishes between physiological fluctuations and pathological abnormalities based on the matching results, including: Obtain the user's real-time blood pressure data at the current moment, and call the corresponding personalized dynamic blood pressure baseline at the current moment to calculate the deviation between the real-time blood pressure and the personalized dynamic blood pressure baseline. Based on the behavioral blood pressure quantification relationship model, the independent influence weights and coupled superimposed influence weights of the user's current time-series behavior and historical time-series behavior, the theoretical fluctuation range of the current blood pressure is predicted. The deviation between the real-time blood pressure and the personalized dynamic blood pressure baseline is compared and matched with the theoretical fluctuation range to obtain the matching result; If the deviation value of the matching result is within the predicted fluctuation range, it is determined to be a physiological fluctuation; if the deviation value exceeds the predicted fluctuation range, it is determined to be a pathological abnormality.

6. The method for determining blood pressure risk based on user behavior status according to claim 1, characterized in that, When a pathological abnormality is identified, the target influence behavior is determined based on the independent influence weights and / or coupled superimposed influence weights, including: When blood pressure is determined to be pathologically abnormal, extract all behaviors that occur within the current time window and their corresponding independent influence weights. and / or; Extract the weights of the coupled and superimposed effects of multiple concurrent behaviors within the same time window; The independent influence weights and the coupled superimposed influence weights are sorted, and the behaviors or combinations of behaviors that have the greatest impact on blood pressure abnormalities are selected based on the sorting results. Determine the preset threshold; Identify behaviors or combinations of behaviors with an impact weight greater than a preset threshold, and determine the behaviors or combinations of behaviors as target impact behaviors.

7. The method for determining blood pressure risk based on user behavior status according to claim 1, characterized in that, Based on the target behavior, output intervention recommendations, including: The target time point and target effect intensity of the behavior that causes abnormal effects on blood pressure are determined based on the type of the target behavior, the time of occurrence, the time of onset of effect, and the time of decay of effect. The intervention priority and intensity for the target behavior are calculated based on the behavioral blood pressure quantification relationship model, the independent influence weights, and the coupled superimposed influence weights. Based on the intervention priority and the intervention intensity, a personalized intervention strategy is generated that matches the target behavior. Transform personalized intervention strategies into quantifiable intervention recommendations; The intervention recommendation information will be output.

8. A blood pressure risk assessment device based on user behavior status, characterized in that, The device includes: The acquisition unit is used to acquire users' time-series behavioral data and time-series blood pressure data; The generation unit is used to annotate the time-series behavioral data with behavioral impact timeliness, marking the time of occurrence of the behavior, the period of onset of effect, and the period of decay of effect, respectively, so as to generate a behavioral time-series feature sequence with timeliness labels; The quantization unit is used to time-align the temporal behavioral feature sequence with the temporal blood pressure data, and quantify the independent influence weight of a single behavior on blood pressure and the coupled superposition influence weight when multiple behaviors occur concurrently through a temporal causal inference model. The first construction unit is used to construct a behavioral blood pressure quantification relationship model based on the independent influence weights and the coupled superimposed influence weights. The second construction unit is used to construct a user-personalized dynamic blood pressure baseline based on normal blood pressure standards, user's historical blood pressure data, and the behavioral blood pressure quantification relationship model. The matching unit is used to match the user's real-time blood pressure data, the personalized dynamic blood pressure baseline, and the behavioral blood pressure quantification relationship model, and to distinguish between physiological fluctuations and pathological abnormalities based on the matching results. The determination unit is used to determine the target influence behavior based on the independent influence weights and / or coupled superimposed influence weights when the abnormality is determined to be pathological. The output unit is used to output intervention suggestion information based on the target influencing behavior.

9. A blood pressure risk assessment device based on user behavior status, characterized in that, The device includes: Processor, memory, input / output units, and bus; The processor is connected to the memory, the input / output unit, and the bus; The memory stores a program, which the processor invokes to perform the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium contains a program that, when executed on a computer, performs the method as described in any one of claims 1 to 7.

Images