A cardiovascular medicine disease early warning analysis system based on big data

Abstract

The application discloses a cardiovascular internal medicine disease early warning analysis system based on big data, relates to the technical field of disease early warning, and comprises a data processing and modeling module, a signal decoupling and correction module, a risk quantification evaluation module and an early warning and adaptive optimization module.The data processing and modeling module is used for collecting physiological signal data and home behavior data of a patient and constructing a physiological baseline model representing the personalized health state of the patient.The signal decoupling and correction module is used for calculating a corrected physiological index stripped of the influence of behavior disturbance based on the real-time collected home behavior data.The risk quantification evaluation module is used for generating a composite risk score representing the current degree of cardiac function deterioration based on the deviation degree of the corrected physiological index from the physiological baseline model.The early warning and adaptive optimization module performs risk level determination and triggers corresponding early warning.The application decouples behavior and physiological signals, constructs a personalized baseline, quantifies multidimensional risks, forms a closed-loop optimization, solves the high false alarm and missed alarm problems of traditional early warning, and realizes precise early warning and long-term effective monitoring.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of disease early warning, and particularly relates to a cardiovascular internal medicine disease early warning analysis system based on big data. BACKGROUND

[0002] As the terminal stage of various cardiovascular diseases, the disease management of patients during the home period after discharge is crucial, especially for the elderly patients living alone. The existing home monitoring system usually collects physiological indicators such as heart rate and blood pressure by means of wearable devices, and provides a basic data collection and preliminary warning means for patient home health monitoring by setting a fixed threshold for early warning.

[0003] The existing home monitoring system faces the core technical problem that the physiological indicator fluctuations caused by the daily activities of patients such as housework and emotional excitement are highly similar to the physiological changes caused by the actual deterioration of heart function in signal, forming a behavior-physiological signal fuzzy coupling effect. The traditional early warning model cannot effectively distinguish between the two kinds of signal fluctuations, resulting in high false positive rate and high false negative rate of early warning, reducing the reliability and timeliness of the system, and it is difficult to meet the real home care needs. SUMMARY

[0004] The purpose of the present application is to provide a cardiovascular internal medicine disease early warning analysis system and method based on big data, which solves the problems existing in the background art.

[0005] To solve the above technical problems, the present application provides a cardiovascular internal medicine disease early warning analysis system based on big data, comprising: a data processing and modeling module for collecting physiological signal data and home behavior data of patients, and based on the physiological signal data in a preset resting steady state, a physiological baseline model representing the personalized health status of the patient is constructed;

[0006] A signal decoupling and correction module is used to identify the current activity state of the patient based on the real-time collected home behavior data, and combine the current activity state with the real-time collected physiological signal data to calculate the corrected physiological indicators stripped of the influence of behavior disturbance through a preset behavior-physiological signal decoupling model;

[0007] A risk quantification evaluation module is used to calculate and generate a standardized physiological deviation index based on the deviation degree of the corrected physiological indicators and the physiological baseline model, and combine a plurality of standardized physiological deviation indexes and a preset clinical weight for weighting to generate a composite risk score representing the current heart function deterioration degree;

[0008] The early warning and adaptive optimization module is configured to determine the risk level and trigger the corresponding early warning based on the composite risk score, and in response to the manual feedback on the early warning, use the real physiological indicators confirmed by the feedback to adaptively optimize the physiological baseline model.

[0009] Preferably, the data processing and modeling module includes the following specific steps:

[0010] S1, calculate the statistical mean of the time series of all physiological signal data collected under the preset resting steady state to generate a personalized mean;

[0011] S2, Calculate the statistical standard deviation of the time series of the physiological signal data to generate a personalized standard deviation;

[0012] S3, the personalized mean and the personalized standard deviation are used together as parameters to define the physiological baseline model.

[0013] Preferably, the signal decoupling and correction module is specifically configured to:

[0014] S21, based on the data collected within the calibration period, calculate the difference between the physiological signal data and the personalized mean in each activity state, and average the difference to generate an average physiological offset corresponding to each activity state.

[0015] S22, subtract the average physiological offset corresponding to the current activity state from the real-time collected physiological signal data to generate the corrected physiological index.

[0016] Preferably, the risk quantification assessment module is specifically configured to generate the standardized physiological deviation index for:

[0017] Calculate the absolute difference between the corrected physiological index and the personalized mean, and divide the absolute difference by the personalized standard deviation to complete the normalization and generate the standardized physiological deviation index.

[0018] Preferably, the risk quantification assessment module, in order to generate the composite risk score, is further configured to:

[0019] Each of the standardized physiological deviation indices is multiplied by a preset clinical weight coefficient to obtain a set of products; then all the products in the set are added together to generate the composite risk score.

[0020] The clinical weighting coefficients are preset based on an expert knowledge base in the field of cardiovascular disease management, or are calibrated by machine learning training on a historical case database.

[0021] Preferably, the early warning and adaptive optimization module is specifically configured to:

[0022] The composite risk score is compared with a preset first risk threshold and a second risk threshold, wherein the second risk threshold is higher than the first risk threshold.

[0023] When the composite risk score is higher than the second risk threshold, it is determined to be a level two warning, and a detailed alarm containing the composite risk score and one or more of the standardized physiological deviation indices that contribute the most to the score is sent to the preset emergency contact.

[0024] When the composite risk score is higher than the first risk threshold but not higher than the second risk threshold, it is determined to be a level one warning, and a voice prompt is issued to the patient through the smart terminal and the collection frequency of the physiological signal data is increased.

[0025] When the composite risk score is not higher than the first risk threshold, the status is determined to be safe, and the current monitoring status is maintained.

[0026] The first risk threshold and the second risk threshold are determined by performing receiver operating characteristic (ROC) curve analysis on historical early warning data to achieve a balance between the sensitivity and specificity of the early warning.

[0027] Preferably, the early warning and adaptive optimization module is configured to adaptively optimize the physiological baseline model and is further configured to:

[0028] When a real physiological indicator confirmed by the human feedback is received, the real physiological indicator is weighted and averaged with the personalized mean to calculate and update the personalized mean.

[0029] Beneficial effects

[0030] Compared with the prior art, the present invention has the following beneficial effects:

[0031] 1. Through the signal decoupling and correction module, the patient's activity status is identified in real time, the disturbance of behavior to physiological signals is removed, and corrected physiological indicators are generated, which fundamentally solves the problem of signal fuzzy coupling, significantly improves the accuracy of early warning, and reduces the false alarm and missed alarm rates.

[0032] 2. With the help of the data processing and modeling module, a personalized physiological baseline model is constructed based on the patient's resting homeostasis data. This model abandons the traditional one-size-fits-all fixed threshold mode and uses the individual as a reference to assess risk. It can keenly capture key physiological changes in specific patients and achieve personalized and precise monitoring.

[0033] 3. The risk quantification assessment module applies clinical weights to the multi-dimensional standardized physiological deviation index to generate a composite risk score. By integrating multi-dimensional information and clinical experience, it intuitively represents the comprehensive risk level of cardiac function deterioration, providing a reliable basis for clinical decision-making.

[0034] 4. The early warning and adaptive optimization module constructs a closed-loop mechanism of "monitoring-early warning-intervention-feedback-optimization". Based on human feedback, it dynamically optimizes the physiological baseline model to continuously adapt to the patient's long-term physiological changes, ensuring the effectiveness and accuracy of the system's long-term monitoring. Attached Figure Description

[0035] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0036] Figure 1 This is a logic block diagram of the system of the present invention. Detailed Implementation

[0037] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0038] Example 1:

[0039] Please see Figure 1 The present invention provides a cardiovascular disease early warning and analysis system based on big data, including: a data processing and modeling module, used to collect patients' physiological signal data and home behavior data, and to construct a physiological baseline model representing the patient's personalized health status based on the physiological signal data under a preset resting steady state;

[0040] The signal decoupling and correction module is used to identify the patient's current activity state based on the real-time collected home behavior data, and combine the current activity state with the real-time collected physiological signal data to calculate the corrected physiological indicators stripped of the influence of behavioral disturbances through a preset behavior-physiological signal decoupling model.

[0041] The risk quantification assessment module is used to calculate and generate a standardized physiological deviation index based on the degree of deviation between the corrected physiological indicators and the physiological baseline model, and to combine multiple standardized physiological deviation indices with preset clinical weights to generate a composite risk score that characterizes the current degree of deterioration of cardiac function.

[0042] The early warning and adaptive optimization module is configured to determine the risk level and trigger the corresponding early warning based on the composite risk score, and in response to the manual feedback on the early warning, use the real physiological indicators confirmed by the feedback to adaptively optimize the physiological baseline model.

[0043] In this embodiment, the system first establishes a unique health baseline for each heart failure patient through a data processing and modeling module. Subsequently, the signal decoupling and correction module, a core component of the invention, accurately identifies and quantifies the temporary impact of daily activities on physiological signals, separating it from the raw data. This effectively solves the problem of false alarms and missed alarms caused by the "fuzzy coupling effect of behavior and physiological signals." The risk quantification assessment module transforms multi-dimensional, pure physiological indicators into a single, clinically significant risk score, providing a quantitative basis for subsequent decision-making. Finally, the early warning and adaptive optimization module forms a complete closed loop of "monitoring-early warning-intervention-feedback-optimization," which not only responds promptly to health risks but also continuously evolves the system model through learning from confirmed data, thereby achieving long-term, accurate, and highly personalized home health monitoring.

[0044] Example 2:

[0045] The data processing and modeling module includes the following specific steps:

[0046] S1, calculate the statistical mean of the time series of all physiological signal data collected under the preset resting steady state to generate a personalized mean;

[0047] S2, Calculate the statistical standard deviation of the time series of the physiological signal data to generate a personalized standard deviation;

[0048] S3, the personalized mean and the personalized standard deviation are used together as parameters to define the physiological baseline model.

[0049] Furthermore, the signal decoupling and correction module is specifically configured for:

[0050] S21, based on the data collected within the calibration period, calculate the difference between the physiological signal data and the personalized mean in each activity state, and average the difference to generate an average physiological offset corresponding to each activity state.

[0051] S22, subtract the average physiological offset corresponding to the current activity state from the real-time collected physiological signal data to generate the corrected physiological index.

[0052] In this embodiment, the data processing and modeling module uses a Gaussian distribution as the physiological baseline model. The design logic of this model is that the physiological indicators of a healthy individual will fluctuate within a certain range around a stable central value under specific conditions. The mathematical expression of the model is as follows:

[0053]

[0054] Indicates the first The true physiological baseline of each physiological indicator;

[0055] The numbering of physiological indicators;

[0056] This represents the Gaussian distribution, i.e., the normal distribution.

[0057] For the first The individualized mean values ​​of each indicator under resting steady state are calculated through step S1.

[0058] For the first The square root of the individual variance of each indicator under resting steady state The result is obtained through step S2.

[0059] Current home monitoring technologies often employ a one-size-fits-all approach based on population statistics, such as uniformly setting a heart rate exceeding 100 beats per minute as abnormal. However, the weakness of this method lies in its complete disregard for individual physiological differences. For an elderly heart failure patient with a consistently high heart rate of 85 beats per minute, a heart rate of 95 beats per minute might be a warning sign; while for an athlete with a baseline heart rate of only 50 beats per minute, 80 beats per minute is perfectly normal. This non-personalized approach leads to underreporting of the former and false positives for the latter.

[0060] To address the aforementioned issues of false alarms and missed alarms, it is necessary to establish a unique reference system for each user that can characterize their own routine health status.

[0061] This invention is inspired by basic medical and statistical theories. In a healthy individual, physiological indicators, under undisturbed resting conditions, do not remain constant but rather exhibit small, approximately random fluctuations around a stable central value. This distribution characteristic statistically closely matches the Gaussian (normal) distribution model. Therefore, this invention employs a Gaussian distribution to establish a mathematical model for each user's physiological indicators.

[0062] The physical significance of the physiological baseline model lies in the fact that it no longer defines "normal" as a fixed value, but rather as a value that includes individualized mean values. ) and personalized fluctuation range ( The "stable state range" of the system. In business terms, this means that the judgment criteria of the early warning system have changed from "whether it exceeds the general red line" to "whether it significantly deviates from its own normal state", thus achieving truly personalized and precise monitoring.

[0063] In the formula, Representing the The true physiological baseline of a physiological indicator (such as heart rate and respiratory rate); It represents a Gaussian distribution.

[0064] (Personalized Mean): This parameter is derived from the calibration phase during the initial deployment of the system. It is obtained by collecting physiological signal data from users over several consecutive days under a preset resting steady state and calculating the arithmetic mean of this time series data (step S1). Its data type is floating-point numerical.

[0065] (Personalized variance): The source of this parameter's variables is related to... The same result is obtained by calculating the statistical variance of the same resting steady-state data (step S2), which represents the natural fluctuation range of this indicator for the user in a healthy state.

[0066] The physiological baseline model constructed here is the cornerstone of the entire early warning analysis system. It serves as the core input for the subsequent risk quantification and assessment module, and is the origin and benchmark for all risk calculations.

[0067] By establishing a physiological baseline model, the system obtains a personalized health profile for each user, abandoning the crude, uniform standard. This technology enables the system to keenly detect subtle but crucial physiological changes that are clinically significant for specific patients, thereby significantly improving the specificity and sensitivity of early warnings in quantitative comparisons.

[0068] The signal decoupling and correction module employs a correction model based on state average offset. The design principle is that the impact of a specific activity on physiological indicators exhibits a certain statistical regularity in the short term, and the true physiological state can be approximated by subtracting the average impact of the activity. The formula for calculating the corrected physiological indicators is as follows:

[0069]

[0070] Indicates time;

[0071] Indicates in Time of the first The corrected physiological value of a physiological indicator is called the corrected physiological indicator.

[0072] Indicates in Time of the first The original observation values ​​of each physiological indicator;

[0073] Indicates in The patient's activity level at any given time, such as resting steady state, daily activity state, or high-intensity transient state.

[0074] Indicates the first Several physiological indicators in active state The average physiological offset is calculated in step S21.

[0075] Current technologies that directly analyze raw physiological signals have a technical drawback: they cannot distinguish between "pathological changes" and "physiological fluctuations." For example, the increased heart rate experienced by a heart failure patient climbing stairs has a signal manifestation that is extremely similar to tachycardia caused by deteriorating cardiac function. This is the so-called "behavioral-physiological signal fuzzy coupling effect." This confusion is the root cause of the high false alarm rate of existing systems.

[0076] In order to accurately assess the true health status, it is necessary to first precisely "strip" away the temporary, benign effects brought about by daily activities from the original physiological signals.

[0077] The design concept of this invention originates from the "noise cancellation" principle in signal processing. If the noise pattern, i.e., the average impact of a specific activity on physiological signals, can be known in advance, a cleaner original signal can be recovered by subtracting this noise pattern from the noisy signal. This invention treats the impact of the activity as a form of physiological noise.

[0078] The physical significance lies in the fact that it estimates in At any time, even if the user is active. Its physiological indicators, after excluding the influence of this activity, are equivalent to the values ​​in the resting state. It is a "pure" physiological indicator that better reflects the user's internal compensatory state of cardiac function.

[0079] Parameter definition:

[0080] : Time of the first Corrected physiological values ​​of each physiological indicator;

[0081] : Raw observations collected by the time sensor;

[0082] Core parameters, representing the activity status. For the The average physiological deviation caused by each physiological indicator.

[0083] Source of variables: The variables originate from step S21. During the calibration period, the system requires the user to wear the device and perform different types of activities, simultaneously collecting physiological data and activity status. The system then calculates the physiological indicators relative to the personalized resting mean for each activity status. The average of the differences is used to obtain a lookup table or model that stores the different values. (e.g., rest, daily activities, high-load transients) corresponding to value.

[0084] Calculated corrected physiological indicators This will serve as the direct input for calculating the standardized physiological deviation index in the next step of the risk quantification assessment module.

[0085] The technological effect is a significant improvement in the accuracy of early warnings. By effectively eliminating activity interference, the system can focus on identifying pathological changes truly caused by deteriorating cardiac function, fundamentally solving the "fuzzy coupling" problem, making the triggering of early warning events more precise, and greatly improving the system's reliability and practical value.

[0086] Example 3:

[0087] The risk quantification assessment module is specifically configured to generate the standardized physiological deviation index for:

[0088] Calculate the absolute difference between the corrected physiological index and the personalized mean, and divide the absolute difference by the personalized standard deviation to complete the normalization and generate the standardized physiological deviation index.

[0089] Furthermore, the risk quantification and assessment module, in order to generate the composite risk score, is further configured to:

[0090] Each of the standardized physiological deviation indices is multiplied by a preset clinical weight coefficient to obtain a set of products; then all the products in the set are added together to generate the composite risk score.

[0091] The clinical weighting coefficients are preset based on an expert knowledge base in the field of cardiovascular disease management, or are calibrated by machine learning training on a historical case database.

[0092] In this embodiment, the risk quantification assessment module first calculates the standardized physiological deviation index. This index is designed using statistical standardization methods to transform physiological indicators with different units and degrees of variability onto the same dimensionless evaluation scale, allowing for direct comparison and integration. Its calculation formula is as follows:

[0093]

[0094] : Time of the first The standardized physiological deviation index of an indicator is a dimensionless floating-point value.

[0095] The corrected physiological values ​​calculated in the previous step;

[0096] Personalized means derived from a physiological baseline model;

[0097] Individualized standard deviation from the physiological baseline model (note that it is) (the square root).

[0098] After obtaining the corrected physiological indicators, directly comparing the absolute values ​​of their deviations from the mean presents new technical challenges. For example, if the heart rate deviates by 10 beats per minute and the respiratory rate deviates by 5 breaths per minute, which deviation is more dangerous? Because their units (dimensions) and inherent fluctuations (degree of variability) are completely different, they cannot be directly compared or added together.

[0099] In order to measure and integrate the risk levels of different physiological indicators on the same scale, it is necessary to standardize the deviation of each indicator.

[0100] This invention explicitly introduces the crucial statistical concept of "Z-score" as its scientific basis. The core idea of ​​the Z-score is to measure how many standard deviations a data point differs from its mean. It is a natural, dimensionless metric, perfectly solving the problem of scaling different indicators.

[0101] The physical meaning of this is that it measures not the "absolute value" of the deviation, but the "significance" or "relative fluctuation" of the deviation. A value of 2.0, whether corresponding to heart rate or respiration, has the same business meaning: "This indicator has deviated from the individual's normal level by two standard deviations." This is a risk measure with universal comparability.

[0102] : Time of the first The standardized physiological deviation index of an indicator is a dimensionless floating-point value. The corrected physiological values ​​calculated in the previous step; Personalized means derived from a physiological baseline model; Individualized standard deviation from the physiological baseline model (note that it is) (the square root).

[0103] Dimensionlessness was ultimately achieved by dividing the numerator (unit: X) by the denominator (unit: X).

[0104] Value range: The value ranges from 0 to 0 for real numbers. The larger the value, the further it deviates from the individual's normal state, and the higher the risk.

[0105] A set of calculations (One for each indicator) will serve as the input for the next step, the "weighted fusion module," to calculate the composite risk score.

[0106] The technological advantage lies in enabling multi-dimensional physiological information to "compete on the same stage." It transforms data with different units and fluctuations into a unified evaluation scale, laying a mathematical foundation for subsequent meaningful weighted fusion and comprehensive assessment, thus making comprehensive risk assessment possible.

[0107] Subsequently, the module calculates a composite risk score using a weighted fusion model. This model is designed to integrate information from multiple physiological dimensions and assign corresponding risk contributions to different indicators based on clinical medical knowledge, thereby generating a comprehensive and singular risk assessment result. The calculation formula is as follows:

[0108]

[0109] For the first A physiological indicator in Standardized physiological deviation index at time;

[0110] , , The definition is the same as above;

[0111] In order to be in The composite risk score at any given moment;

[0112] The total number of core physiological indicators used for risk assessment is a systematic constant pre-defined according to clinical guidelines.

[0113] For the first The clinical weight coefficients of each indicator are preset by a medical expert knowledge base or obtained through machine learning training using historical case data.

[0114] Complex diseases like heart failure often result from the combined effects of multiple physiological systems. Monitoring only a single indicator can lead to oversimplification and fail to comprehensively reflect the overall risk.

[0115] A model is needed that can integrate the risk information contained in multiple "standardized physiological deviation indices" into a single total score that can comprehensively represent the overall risk level of current cardiac function deterioration.

[0116] Weighted summation is a classic method in multi-attribute decision theory and index construction. Its technical implication lies in its ability to intuitively reflect the importance of different factors in the final decision. This invention draws on this idea to assign corresponding weights to the risk contribution of different physiological indicators.

[0117] The business significance lies in the fact that it no longer presents doctors or users with a bunch of scattered data, but instead provides an intuitive, quantifiable snapshot of a single risk. For example, A score of 7.5 is more dangerous than 3.2, and this score is directly related to clinical decision-making.

[0118] Weighting coefficient This is the core of the formula, and its value is not set arbitrarily. This invention provides a clear method for determining it:

[0119] When the system is initialized or lacks sufficient historical data, the weights can be set empirically by a group of cardiovascular experts based on clinical guidelines. For example, for patients with heart failure, the weights for indicators such as dyspnea and nocturnal paroxysmal dyspnea can be determined. It will have a higher weight than heart rate variability.

[0120] Machine Learning Training: A more advanced and accurate approach involves using massive amounts of labeled historical case data for supervised learning training using machine learning algorithms such as logistic regression, support vector machines (SVM), or gradient boosting trees. The algorithm will automatically optimize to find a set of weights that maximizes the accuracy of the predictive model. .

[0121] The weight coefficients are designed to be configurable in the background, and the system is adaptive, able to periodically retrain the model using newly accumulated case data and automatically update the weights to adapt to the development of disease understanding or changes in specific patient groups.

[0122] Calculated composite risk score This will be directly used for the next step of threshold determination.

[0123] The system presets at least two levels of risk thresholds (a first risk threshold and a second risk threshold), by... By comparing these thresholds, the system can automatically classify "safe status", "level 1 warning" and "level 2 warning".

[0124] Compared with existing technologies that provide fragmented indicators, the present invention provides... By quantitatively comparing multi-dimensional information and incorporating clinical experience and knowledge, the technology outputs a more comprehensive, intuitive, and information-dense integrated risk level, providing a more reliable and integrated basis for subsequent precise classification and early warning and clinical decision-making.

[0125] Example 4:

[0126] The early warning and adaptive optimization module is specifically configured for:

[0127] The composite risk score is compared with a preset first risk threshold and a second risk threshold, wherein the second risk threshold is higher than the first risk threshold.

[0128] When the composite risk score is higher than the second risk threshold, it is determined to be a level two warning, and a detailed alarm containing the composite risk score and one or more of the standardized physiological deviation indices that contribute the most to the score is sent to the preset emergency contact.

[0129] When the composite risk score is higher than the first risk threshold but not higher than the second risk threshold, it is determined to be a level one warning, and a voice prompt is issued to the patient through the smart terminal and the collection frequency of the physiological signal data is increased.

[0130] When the composite risk score is not higher than the first risk threshold, the status is determined to be safe, and the current monitoring status is maintained.

[0131] The first risk threshold and the second risk threshold are determined by performing receiver operating characteristic (ROC) curve analysis on historical early warning data to achieve a balance between the sensitivity and specificity of the early warning.

[0132] Furthermore, the early warning and adaptive optimization module is further configured to adaptively optimize the physiological baseline model and is used for:

[0133] When a real physiological indicator confirmed by the human feedback is received, the real physiological indicator is weighted and averaged with the personalized mean to calculate and update the personalized mean.

[0134] In this embodiment, the risk level determination function of the early warning and adaptive optimization module is based on the composite risk score. The threshold for implementation is typically determined through receiver operating characteristic (ROC) curve analysis of historical data to balance the sensitivity and specificity of the warning. Its adaptive optimization function employs an exponentially weighted moving average algorithm. The design logic of this algorithm is to fine-tune the old baseline mean using newly confirmed real data, ensuring the smoothness of model updates through a small learning rate factor. This allows the model to slowly adapt to the patient's long-term physiological changes, rather than reacting drastically to a single feedback. The formula for calculating the updated personalized mean is as follows:

[0135]

[0136] The updated version New personalized mean values ​​for each indicator;

[0137] The old personalized mean before the update;

[0138] The learning rate factor is a pre-defined, relatively small constant.

[0139] The key input is not arbitrary real-time data, but rather the actual physiological state indicators confirmed by medical staff or professional family members after the system triggers an alert, through manual intervention and verification. This is a validated, high-quality feedback data set.

[0140] Existing technological limitations / pain points: Most existing monitoring models are static, remaining unchanged once deployed. However, the human body's health status is dynamically evolving. Whether due to the natural progression of disease over time, aging, or improvement due to effective treatment, the physiological baseline undergoes long-term, slow changes. Static models will therefore gradually become inaccurate, leading to a decline in accuracy.

[0141] A closed-loop feedback mechanism is needed to enable the physiological baseline model to "self-evolve" and "continuously learn," thereby dynamically adapting to changes in users' long-term health trends and maintaining the model's long-term effectiveness.

[0142] This invention draws inspiration from the classic "exponentially weighted moving average" algorithm in time series analysis. The core idea of ​​EWMA is to assign higher weights to recent data and exponentially decaying weights to older data, thereby smoothing short-term fluctuations while effectively tracking long-term trends.

[0143] Physical meaning: The physical meaning of this formula is to make the individual mean Towards a "new true value" confirmed by human feedback from professionals. This allows for small, gradual adjustments. It ensures that the model's baseline slowly and steadily follows the patient's actual long-term physiological changes, rather than reacting drastically to a single feedback.

[0144] Determining the learning rate: This is the key adjustable parameter in this formula, with a value range between (0,1). Its selection involves a trade-off: a smaller value indicates a higher risk of falling below 0. This means slow updates, a very stable model, and the ability to effectively filter out accidental and noisy feedback, focusing more on long-term trends; larger... This means that the model is more sensitive to new feedback and can adapt to changes more quickly.

[0145] In practice, The value is typically set as an empirical value based on clinical needs. For example, for the management of chronic diseases with slow disease progression, a very small value might be chosen to ensure model stability. This value can also be designed as a configurable option in the backend.

[0146] This update process constitutes a complete closed-loop management system of "monitoring-early warning-intervention-feedback-optimization". This update process is triggered when the system receives human feedback.

[0147] Compared to static models, the adaptive optimization module designed in this invention is one of its core advancements. Its technical effect is to endow the system with the ability to self-evolve and continuously optimize. The model can keep pace with the patient's long-term health trends, ensuring continuous effectiveness and accuracy throughout the entire long-term monitoring period, thus achieving truly dynamic and intelligent health management.

[0148] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A cardiovascular disease early warning and analysis system based on big data, characterized in that, include: The data processing and modeling module is used to collect the patient's physiological signal data and home behavior data, and to construct a physiological baseline model representing the patient's personalized health status based on the physiological signal data under a preset resting steady state. The signal decoupling and correction module is used to identify the patient's current activity state based on the real-time collected home behavior data, and combine the current activity state with the real-time collected physiological signal data to calculate the corrected physiological indicators stripped of the influence of behavioral disturbances through a preset behavior-physiological signal decoupling model. The risk quantification assessment module is used to calculate and generate a standardized physiological deviation index based on the degree of deviation between the corrected physiological indicators and the physiological baseline model, and to combine multiple standardized physiological deviation indices with preset clinical weights to generate a composite risk score that characterizes the current degree of deterioration of cardiac function. The early warning and adaptive optimization module is configured to determine the risk level and trigger the corresponding early warning based on the composite risk score, and in response to the manual feedback on the early warning, use the real physiological indicators confirmed by the feedback to adaptively optimize the physiological baseline model. The data processing and modeling module includes the following specific steps: S1, calculate the statistical mean of the time series of all physiological signal data collected under the preset resting steady state to generate a personalized mean; S2, Calculate the statistical standard deviation of the time series of the physiological signal data to generate a personalized standard deviation; S3, the personalized mean and the personalized standard deviation are used together as parameters to define the physiological baseline model; The signal decoupling and correction module is specifically configured to: S21, based on the data collected within the calibration period, calculate the difference between the physiological signal data and the personalized mean in each activity state, and average the difference to generate an average physiological offset corresponding to each activity state. S22, subtract the average physiological offset corresponding to the current activity state from the real-time collected physiological signal data to generate the corrected physiological index.

2. The cardiovascular disease early warning and analysis system based on big data according to claim 1, characterized in that, The risk quantification assessment module is specifically configured to generate the standardized physiological deviation index for: Calculate the absolute difference between the corrected physiological index and the personalized mean, and divide the absolute difference by the personalized standard deviation to complete the normalization and generate the standardized physiological deviation index.

3. The cardiovascular disease early warning and analysis system based on big data according to claim 2, characterized in that, The risk quantification and assessment module, in order to generate the composite risk score, is further configured to: Each of the standardized physiological deviation indices is multiplied by a preset clinical weight coefficient to obtain a set of products; then all the products in the set are added together to generate the composite risk score. The clinical weighting coefficients are preset based on an expert knowledge base in the field of cardiovascular disease management, or are calibrated by machine learning training on a historical case database.

4. The cardiovascular disease early warning and analysis system based on big data according to claim 3, characterized in that, The early warning and adaptive optimization module is specifically configured to: The composite risk score is compared with a preset first risk threshold and a second risk threshold, wherein the second risk threshold is higher than the first risk threshold. When the composite risk score is higher than the second risk threshold, it is determined to be a level two warning, and a detailed alarm containing the composite risk score and one or more of the standardized physiological deviation indices that contribute the most to the score is sent to the preset emergency contact. When the composite risk score is higher than the first risk threshold but not higher than the second risk threshold, it is determined to be a level one warning, and a voice prompt is issued to the patient through the smart terminal and the collection frequency of the physiological signal data is increased. When the composite risk score is not higher than the first risk threshold, the status is determined to be safe, and the current monitoring status is maintained. The first risk threshold and the second risk threshold are determined by performing receiver operating characteristic (ROC) curve analysis on historical early warning data to achieve a balance between the sensitivity and specificity of the early warning.

5. The cardiovascular disease early warning and analysis system based on big data according to claim 4, characterized in that, The early warning and adaptive optimization module is used to adaptively optimize the physiological baseline model and is further configured to: When a real physiological indicator confirmed by the human feedback is received, the real physiological indicator is weighted and averaged with the personalized mean to calculate and update the personalized mean.

Images