Management system for informatization planning of crowd health archives

Abstract

The invention relates to the technical field of medical data processing and health information management, in particular to an informatization planning crowd health archive management system comprising an archive data aggregation module used for collecting multi-dimensional health records and calling a knowledge graph; the physiological reference reconstruction module is used for constructing an ideal physiological reference; the pathological disturbance simulation module is used for generating a pathological simulation state; the double-track differential extraction module is used for respectively calculating deviations between the time sequence physical sign data and the ideal benchmark and between the pathological simulation state and the ideal benchmark, and obtaining real and theoretical residual error sequences; the coupling research and judgment module is used for comparing the waveform form similarity of the residual sequence to judge the abnormal property; the dynamic intervention module is used for generating a graded intervention instruction if the effective health risk is judged, and executing data filtering if the environment noise is judged; the system further comprises a self-adaptive correction module which is used for finely adjusting construction parameters of the ideal physiological reference by using a non-pathological sample; according to the invention, the false positive alarm rate is reduced, and accurate delivery of medical resources is ensured.

Description

Technical Field

[0001] This invention relates to the field of medical data processing and health information management technology, specifically to an information-based population health record management system. Background Technology

[0002] With the rapid development of medical information technology, the physiological data dimensions covered by population health records have shown explosive growth. The complexity of this multi-source heterogeneous data has brought severe challenges to the continuous tracking and accurate assessment of health status, especially in the automated analysis of massive time-series vital signs data. Currently, traditional health data management and monitoring methods mainly rely on preset static numerical thresholds for alarms, or only archive physical examination records at a single moment. However, existing technologies lack first-principles modeling of an individual's natural physiological decline trajectory, making it difficult to effectively isolate environmental noise interference caused by factors such as differences in measurement posture and short-term emotional fluctuations. This judgment logic that relies solely on the absolute magnitude of values ​​cannot distinguish between benign physiological fluctuations and real pathological evolution from waveform morphology, resulting in a high false positive alarm rate and easy to miss hidden early pathological signals. Therefore, how to accurately filter out noise and identify effective health risks in complex dynamic physiological data has become an urgent problem to be solved in this field. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention provides an information-based population health record management system. Specifically, the technical solution of this invention includes: The archive data aggregation module is used to collect historical health records of the target object from multiple dimensions, obtain time-series vital sign data and static physiological attribute data, and retrieve preset medical knowledge graphs; The physiological baseline reconstruction module is used to construct the dynamic health trajectory of the target object under ideal and compliant conditions based on static physiological attribute data and medical knowledge graph, which serves as the ideal physiological baseline. The pathological perturbation simulation module is used to define pathological perturbation functions based on medical knowledge graphs and superimpose the pathological perturbation functions onto ideal physiological benchmarks to generate pathological simulation states that include theoretical pathological features. The dual-track differential extraction module is used to calculate the deviation between time-series vital sign data and ideal physiological benchmarks to obtain the actual residual sequence, and to calculate the deviation between the pathological simulation state and ideal physiological benchmarks to obtain the theoretical residual sequence. The coupling analysis module is used to compare the waveform morphology of the actual residual sequence and the theoretical residual sequence, calculate the morphological similarity, and determine the abnormal nature of the time series vital signs data based on the morphological similarity. The dynamic intervention module is used to generate tiered intervention instructions if the abnormality is determined to be a valid health risk, and to perform data filtering operations if the abnormality is determined to be environmental noise.

[0004] Preferably, the physiological baseline reconstruction module is specifically used for: Analyze static physiological attribute data to determine the age and gender parameters of the target object; Based on age and gender parameters, the corresponding natural physiological decline curves are matched from the medical knowledge graph; Based on the natural physiological decline curve and the steady-state rhythm function in the medical knowledge graph, a physiological steady-state differential equation containing the metabolic rate decline variable and the natural physiological drift rate variable is established. Solve the physiological homeostasis differential equation to generate time series curves that do not contain pathological mutation characteristics, which serve as an ideal physiological benchmark.

[0005] Preferably, the pathological disturbance simulation module is specifically used for: Extract pathological causes and violation patterns from medical knowledge graphs and transform them into calculable perturbation factors; A set of pathological perturbation functions is constructed, which includes a concentration drop function for simulating drug adherence failure and a peak decay coefficient for simulating metabolic function deterioration. The pathological perturbation functions in the set of pathological perturbation functions are injected into the ideal physiological benchmark to generate multiple hypothetical pathological curves, and each hypothetical pathological curve is used as a pathological simulation state.

[0006] Preferably, the dual-track differential extraction module is specifically used for: For each pathological simulation state, a difference operation with the ideal physiological benchmark is performed to generate a pure pathological deviation waveform; The morphological features of the pure pathological deviation waveform are extracted, the influence of baseline drift is removed, and the processed waveform data is used as the theoretical residual sequence, which characterizes the theoretical deviation morphology caused by a specific pathological cause.

[0007] Preferably, the coupling analysis module is specifically used for: Time axis alignment is performed between the actual residual sequence and the theoretical residual sequence; The dynamic time warping algorithm is used to calculate the optimal waveform matching path between the actual residual sequence and the theoretical residual sequence in time series. The Euclidean distance or cosine similarity between two sequences is calculated based on the best matching path, and the calculation results are normalized to obtain the morphological similarity.

[0008] Preferably, when determining the anomalous nature of time-series vital sign data, the coupling analysis module is specifically used for: A preset set of judgment thresholds, which includes risk confirmation thresholds; Compare morphological similarity with risk confirmation thresholds; If the morphological similarity is greater than or equal to the risk confirmation threshold, the actual residual sequence is determined to have pathological evolution characteristics, the abnormal nature is determined to be an effective health risk, and a graded intervention instruction is triggered. If the morphological similarity is less than the risk confirmation threshold, the actual residual sequence is determined to be an occasional fluctuation, and the anomaly is identified as environmental noise.

[0009] Preferably, the dynamic intervention module is specifically used for: Obtain the pathological simulation state corresponding to the determination of a valid health risk, and identify the type of pathological perturbation function associated with the pathological simulation state; Based on the pathological perturbation function type, the corresponding clinical intervention path is retrieved from the medical knowledge graph; Based on the deviation of time-series vital signs data, the risk level is determined, and corresponding graded intervention instructions are generated in conjunction with the clinical intervention pathway.

[0010] Preferably, the archive data aggregation module is specifically used for: Receive raw vital signs signals from multiple heterogeneous monitoring devices; The original vital signs signals are imputed for missing values ​​and smoothed for outliers to generate standardized time-series vital signs data. The time series vital signs data are periodically decomposed to separate the long-term trend component and the seasonal fluctuation component, and the standardized time series vital signs data after noise removal is used as the main input data for subsequent processing.

[0011] Preferably, the system further includes an adaptive correction module, used for: If the abnormality is determined to be environmental noise, the corresponding time-series vital signs data will be marked as non-pathological samples. The parameters for constructing an ideal physiological baseline are fine-tuned using non-pathological samples to update the personalized physiological homeostasis model of the target subjects.

[0012] Compared with the prior art, the present invention has the following beneficial effects: 1. The beneficial effects of this system are as follows: Through the collaborative work of the dual-track differential extraction module and the coupled judgment module, the system performs two independent differential operations to obtain the actual residual sequence and the theoretical residual sequence. It no longer relies solely on numerical magnitude for alarms, but determines the nature of anomalies by calculating the similarity of waveform morphology. This effectively eliminates the interference of environmental noise caused by improper measurement posture or short-term emotional fluctuations on health assessment. Especially in the context of chronic disease management, the dynamic time warping algorithm is used to handle the time asynchrony problem between actual vital sign data and theoretical models, tolerating the deviation of drug onset time due to individual differences. This elevates health risk assessment from traditional threshold monitoring to the dimension of topological trend assessment, significantly reducing the false positive alarm rate and ensuring accurate capture of effective health risks in complex interference environments. 2. This system establishes a comparative benchmark based on physiological homeostasis mechanisms through a physiological benchmark reconstruction module. Utilizing physiological homeostasis differential equations that include metabolic rate decay variables and natural physiological drift rate variables, and combining them with natural physiological decline curves, an ideal physiological benchmark is constructed. This allows interference from natural aging to be eliminated from the dynamic trajectory, achieving accurate simulation of individual physiological homeostasis mechanisms. On this basis, the pathological perturbation simulation module transforms pathological causes in the medical knowledge graph into calculable perturbation factors. By injecting drug concentration drop functions or peak decay coefficients, various hypothetical pathological simulation states are generated, enabling the system to actively guess the pathological path and identify complex data forms that conform to specific pathological characteristics. This allows the system to capture hidden pathological signals before patients show obvious clinical symptoms. 3. This system periodically decomposes the original vital signs signals through the archival data aggregation module. While filtering out high-frequency noise and fixing seasonal fluctuations, it retains long-term trend components above the expected pathological event timescale by setting trend smoothing window parameters, ensuring the pathological information content of the input data. In the dynamic intervention stage, based on the morphological similarity judgment results, after confirming the effective health risk, the system further identifies the associated pathological perturbation function type and calculates the deviation magnitude. Combined with the mean absolute percentage error, it determines the risk level and finally generates graded intervention instructions containing specific action guidelines based on the medical knowledge graph. This realizes a closed-loop auxiliary decision-making process from abnormality detection and explanation of pathological causes to providing targeted countermeasures, improving the feasibility and accuracy of clinical intervention. 4. This system achieves model self-evolution and parameter optimization through an adaptive correction module. When anomalies are identified as environmental noise, the system automatically marks the data as non-pathological samples and uses gradient descent to fine-tune the construction parameters of the ideal physiological benchmark, updating the personalized physiological homeostasis model of the target object. Furthermore, the system strictly adheres to physical dimension consistency during parameter updates, avoiding mathematical logic errors. Simultaneously, during the dual-track differential extraction process, the system performs baseline drift removal on the pathological simulation state, ensuring that the theoretical residual sequence contains only communication morphological features caused by specific pathological reasons. This guarantees the mathematical consistency between theoretical and real data in subsequent waveform comparisons, thereby improving the system's robustness and adaptability during long-term operation. Attached Figure Description

[0013] The present invention will be further explained below with reference to the accompanying drawings and embodiments: Figure 1 This is a structural diagram of the system of the present invention. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0015] Example 1: Please see Figure 1 An information-based population health record management system, including: The archive data aggregation module is used to collect historical health records of the target object from multiple dimensions, obtain time-series vital sign data and static physiological attribute data, and retrieve preset medical knowledge graphs; The physiological baseline reconstruction module is used to construct the dynamic health trajectory of the target object under ideal and compliant conditions based on static physiological attribute data and medical knowledge graph, which serves as the ideal physiological baseline. The pathological perturbation simulation module is used to define pathological perturbation functions based on medical knowledge graphs and superimpose the pathological perturbation functions onto ideal physiological benchmarks to generate pathological simulation states that include theoretical pathological features. The dual-track differential extraction module is used to calculate the deviation between time-series vital sign data and ideal physiological benchmarks to obtain the actual residual sequence, and to calculate the deviation between the pathological simulation state and ideal physiological benchmarks to obtain the theoretical residual sequence. The coupling analysis module is used to compare the waveform morphology of the actual residual sequence and the theoretical residual sequence, calculate the morphological similarity, and determine the abnormal nature of the time series vital signs data based on the morphological similarity. The dynamic intervention module is used to generate tiered intervention instructions if the abnormality is determined to be a valid health risk, and to perform data filtering operations if the abnormality is determined to be environmental noise.

[0016] This embodiment details the overall architecture and collaborative logic of the information-based population health record management system. This system aims to address the problem in existing health management technologies that cannot effectively distinguish between benign physiological fluctuations and actual pathological deterioration. The record data aggregation module serves as the data entry point, collecting multi-dimensional data from the target object's historical health records through a high-frequency sampling interface. This data includes time-series vital sign data such as continuous blood pressure and dynamic blood glucose, as well as static physiological attribute data such as age and gender. Simultaneously, it retrieves a pre-defined medical knowledge graph containing clinical treatment guidelines. The medical knowledge graph is a heterogeneous information network digitally constructed based on publicly available authoritative clinical treatment guidelines, pharmacopoeia data, and publicly available datasets from large-scale epidemiological surveys. Specifically, this knowledge graph uses a triplet topology structure of entity-relationship-attribute for storage. Entities include physiological indicators, intervention factors, demographic labels, and pathological patterns; Edges: Define the logical relationships between entities, including evolutionary relationships that characterize changes with age and induced relationships that characterize drug action mechanisms, etc. Attributes: Stores the numerical physical constraints attached to the entity node; Based on this structure, the system's logic for obtaining key parameters is as follows: Obtaining the discrete point set of the natural physiological decline curve: The system uses physiological indicator entities as anchor points, traverses the associated nodes under different demographic labels through evolutionary association edges, extracts the statistical mean in the attribute slots, and thus constructs a set of discrete points like... The discrete set of points, where, : Derived from the time index node of the demographic tag entity in the medical knowledge graph, representing the first time dimension of the natural physiological decline curve. Each sampling observation point, with dimensions of This parameter is not a continuous-time variable, but a discrete age marker determined by segmented statistics based on large-scale epidemiological survey data, such as... This corresponds to the attribute value constraints of entity nodes in the graph; : The population statistical mean derived from the entity attribute slots of physiological indicators, representing the population statistical mean in the physiological indicator entity attribute slots. At the corresponding age point, the baseline values ​​of standard physiological indicators for the target group's gender and demographic characteristics, in units of [missing information]. This value was obtained by traversing the nodes connected by the evolutionary association edges in the knowledge graph and extracting the pre-stored large sample statistical expectation values ​​from the attribute slots. It is used to subsequently construct a continuous natural physiological decline curve. ; Obtaining standard pharmacokinetic parameters: The system locates a specific intervention factor entity and directly reads the pre-stored half-life constant and standard maximum effect value in its attribute slot; The physiological benchmark reconstruction module establishes a first-principles comparison benchmark. Based on static physiological attribute data and medical knowledge graph, it constructs the dynamic health trajectory of the target object under ideal compliance state. Ideal compliance state refers to the theoretical thermodynamic steady state in which the target object's physiological functions are regulated only by natural biological rhythms and natural physiological aging laws, without any pathological mutation factors or environmental noise interference, under the premise of strictly following the established medical intervention plan. Based on this, the pathological perturbation simulation module generates a hypothetical lesion template, defines a pathological perturbation function based on a medical knowledge graph, and superimposes this function onto an ideal physiological benchmark to generate a pathological simulation state containing theoretical pathological features. Then, the dual-track difference extraction module, as the core computing engine, performs two independent difference operations: on the one hand, it calculates the deviation between the actually collected time-series vital signs data and the ideal physiological benchmark to obtain the actual residual sequence; on the other hand, it calculates the deviation between the pathological simulation state and the ideal physiological benchmark to obtain the theoretical residual sequence. The coupling judgment module compares the waveform morphology of the actual residual sequence and the theoretical residual sequence, calculates the morphological similarity, and no longer relies solely on numerical magnitude for alarms, but judges whether the actual deviation morphology is topologically similar to the theoretically derived lesion morphology. Based on the judgment results, the dynamic intervention module generates graded intervention instructions in response to anomalies deemed as effective health risks; and performs data filtering operations in response to anomalies deemed as environmental noise. In the context of chronic disease management, this embodiment effectively eliminates the interference of environmental noise caused by improper measurement posture or short-term emotional fluctuations on health assessment through a dual-track differential and morphological coupling mechanism. The system utilizes synthetic analysis to elevate health risk assessment from traditional threshold monitoring to trend analysis, significantly reducing the false positive alarm rate and ensuring the accurate allocation of medical resources.

[0017] Example 2: The physiological baseline reconstruction module is specifically used for: Analyze static physiological attribute data to determine the age and gender parameters of the target object; Based on age and gender parameters, the corresponding natural physiological decline curves are matched from the medical knowledge graph; Based on the natural physiological decline curve and the steady-state rhythm function in the medical knowledge graph, a physiological steady-state differential equation containing the metabolic rate decline variable and the natural physiological drift rate variable is established. Solve the physiological homeostasis differential equation to generate time series curves that do not contain pathological mutation characteristics, which serve as an ideal physiological benchmark.

[0018] This embodiment further defines the execution logic of the physiological benchmark reconstruction module. Its core lies in using differential equations to simulate the physiological homeostasis mechanism of the human body; the system analyzes static physiological attribute data to determine the age parameter of the target object. With gender parameter Based on these two parameters, the corresponding natural physiological decline curve is matched from the medical knowledge graph; here, the physical definition and data structure of the curve are clarified: the natural physiological decline curve is pre-stored in the medical knowledge graph. It describes the trajectory of absolute values ​​of physiological indicators with age, rather than the rate of decline; the curve is generated based on large-scale cross-sectional epidemiological survey data and stored in piecewise function form; during matching, the system determines the gender. A family of index functions is used to determine the baseline of the physiological trajectory to which the individual belongs. Based on the natural physiological decline curve and the steady-state rhythm function in the medical knowledge graph, a first-order linear non-homogeneous differential equation is established to describe the changes in physiological indicators under ideal conditions, as shown in the following formula: in, :Time derived from numerical calculation Ideal physiological index values, dimensions ; here The term "physical unit symbol" is used generally, and its specific form depends on the type of physiological indicator being processed by the current system. This designation aims to give the equation a universal adaptability to multi-source heterogeneous physiological data. For example, when the physiological indicator is blood pressure, the dimensionless symbol is used. It is millimeters of mercury (mmHg); Steady-state rhythm function derived from medical knowledge graph, dimensionless This function represents a standard periodic fluctuation template in the absence of pathological conditions, such as circadian rhythms; Stored in the system as a length equal to One-dimensional discrete vector For any time The system calculates its relative index over a 24-hour period: in, Sampling frequency, unit: It should be consistent with the monitoring equipment or data acquisition system; for The length of the sample typically corresponds to the number of sampling points within a complete physiological cycle, i.e. ;symbol The modulo operator, in its physical sense, represents a periodic mapping in the time domain; this operation aims to convert time-varying data into a periodic mapping. The cumulative total number of sampling points is mapped back to the interval. This transforms the monotonically increasing absolute linear time into relative physiological cycle moments, ensuring that the system can cyclically address steady-state rhythm function vectors of finite length. The ideal circadian rhythm value for that moment is obtained through a table lookup operation. : Metabolic rate decay variable derived from parameter analysis, dimensionless This characterizes the body's ability to self-regulate and restore homeostasis. The natural physiological drift rate function, derived from the natural physiological decline curve, is used to strictly conform to the left side of the differential equation. Dimensions This embodiment clearly states The physical dimensions are This refers to the drift per unit time, not the acceleration; the calculation logic is as follows: Differentiation yields the macroscopic rate: for the matched natural physiological decline curve Taking the first derivative with respect to age yields the macroscopic rate of decline, whose dimensions are... : Dimensional conversion: Introducing a time conversion factor This converts the macroscopic rate into a drift rate on a microscopic monitoring timescale: At this point, the dimensions are converted to ; Fitting the time-varying function: due to the monitoring period Over time A slight increase, the system calculates the corresponding value within the current monitoring window. The sequence was fitted with a time-varying value. Quadratic function: The physical meaning of this correction factor is in terms of matching rate dimensions: ,dimension This serves as the reference drift rate for monitoring the initial moment; ,dimension , is the linear rate of change of the drift rate, i.e., the acceleration of physiological decline; ,dimension , is the second rate of change of the drift velocity; the above coefficients By monitoring the micro decay rate within the monitoring time window The least squares fitting method was used to establish the... A direct mathematical mapping relationship with the physiological decline curve; this definition eliminates the need to... The misuse of the dimensionless acceleration ensures that all terms in the differential equation are velocity terms. The above equations are solved using the Runge-Kutta numerical integration method, with initial conditions set. Generate ideal physiological benchmarks .

[0019] Example 3: The pathological perturbation simulation module is specifically used for: Extract pathological causes and violation patterns from medical knowledge graphs and transform them into calculable perturbation factors; A set of pathological perturbation functions is constructed, which includes a concentration drop function for simulating drug adherence failure and a peak decay coefficient for simulating metabolic function deterioration. The pathological perturbation functions in the set of pathological perturbation functions are injected into the ideal physiological benchmark to generate multiple hypothetical pathological curves, and each hypothetical pathological curve is used as a pathological simulation state.

[0020] This embodiment further defines the execution logic of the pathological perturbation simulation module, aiming to transform medical knowledge into computable mathematical perturbations; the system extracts pathological causes and violation behavior patterns from the medical knowledge graph and transforms them into perturbation factors; Constructing a set of pathological perturbation functions This includes concentration drop functions used to simulate drug adherence failure. Its definition is as follows: This function is used to describe the pathological process in which the concentration or effect of a drug in the body decreases exponentially over time after the occurrence of drug adherence failure. in, The maximum deviation from the pharmacodynamic model is calculated using the following formula: In the formula, The standard maximum effect value pre-stored for this type of drug in the medical knowledge graph, such as the decrease in systolic blood pressure. ; The current prescription dose for the target object is obtained based on electronic medical records or prescription records collected in the archive data aggregation module; This is the standard reference dose; This refers to the individual sensitivity correction coefficient fitted based on historical data of the target object, with a value range of... to The physical meaning is the peak value of the data drop caused by drug failure; its specific calculation logic is as follows: retrieve the historical medication records of the target object. Set of measured maximum effect values ​​during the onset of action of drugs of the same class ,in, This section specifies the number of times medication was used in historical records. The rule for determining the value is: count the number of times all records meeting the criteria for the target object within the last 6 months; if there are fewer than 3 records, then... Forced value to 0; subscript Indicates the first Record of each medication administration; calculate the arithmetic mean of the relative deviations as an initial estimate: If the target individual has no relevant historical medication records, i.e. Then set directly ; To prevent interference from outliers, a truncation function is used for constraints: : A constant derived from pharmacokinetics, its physical meaning is the drug elimination rate constant, and its unit is... Its value is related to the drug's metabolic half-life. Satisfying Relationships ; : Continuous-time variable; The time derived from the assumed conditions, in physical terms, is the assumed time when the missed medication will occur; The unit step function is used to limit the start time of the disturbance, ensuring that... The time perturbation is zero; Simultaneously, a peak decay coefficient is defined to simulate the deterioration of metabolic function. The coefficient This value is derived from the deviation of the target individual's historical glycated hemoglobin (HbA1c) level from normal physiological values. Physically, it represents the proportion of insulin sensitivity loss, and typically falls within a certain range. to In specific calculations, the system uses a linear mapping formula to determine... : in, These are measured values. The upper limit of normal physiological function; constants in the formula This coefficient is a mapping coefficient based on clinical data fitting, used to... Deviation is converted into attenuation ratio; During the calculation process, the system uses the sliding window averaging method to extract the baseline component of the ideal physiological benchmark. The time width of the sliding window is set to It covers approximately 3-5 cardiac cycles to smooth short-term fluctuations, thereby reducing the peak attenuation coefficient. The specific injection formula is as follows: This formula applies to the baseline-removed pulsating component. This operation ensures that the underlying physiological level of the signal is not altered when simulating the blunting and peak suppression characteristics of the blood glucose peak response caused by insulin resistance, thus achieving precise shaping of the peak morphology; the aforementioned pathological perturbation functions are then injected into the ideal physiological baseline. , generate the first Pathological simulation state For the concentration drop function, additive injection is used, i.e. For peak attenuation, the above formula shall be used; In this embodiment, under complex pathological simulation scenarios, the system has the ability to actively guess by using a parameterized pathological perturbation injection mechanism. By simulating multiple possible disease deterioration paths, the system can identify complex data patterns that conform to specific pathological characteristics, thereby capturing hidden pathological signals before the patient shows obvious clinical symptoms.

[0021] Example 4: The dual-track differential extraction module is specifically used for: For each pathological simulation state, a difference operation with the ideal physiological benchmark is performed to generate a pure pathological deviation waveform; The morphological features of the pure pathological deviation waveform are extracted, the influence of baseline drift is removed, and the processed waveform data is used as the theoretical residual sequence, which characterizes the theoretical deviation morphology caused by a specific pathological cause.

[0022] This embodiment further defines the calculation logic of the second track in the dual-track differential extraction module; for each generated pathological simulation state The system performs in accordance with its ideal physiological benchmark. Subtraction operation generates a pure pathological deviation waveform. The formula is: To ensure mathematical consistency between theoretical and real-world data in subsequent comparisons, the system explicitly defines an algorithm for removing baseline drift, namely, performing zero-mean normalization: the time length for pathological simulation calculations is defined as... Calculate the sequence In terms of time length arithmetic mean within And perform the operation: Subscript The range of values ​​is , Represents the total number of generated pathological simulation states; The necessity of this operation lies in the fact that the subsequent coupling and judgment module, i.e., the embodiment, adopts Z-Score normalization, which includes a mean subtraction operation. If the theoretical sequence retains the DC component generated by the pathological function integral, it will lead to morphological alignment distortion. Therefore, this step ensures that the output theoretical residual sequence is accurate. It only includes communication morphological features caused by specific pathological reasons, thereby achieving pure waveform topology comparison.

[0023] Example 5: The coupling analysis module is specifically used for: Time axis alignment is performed between the actual residual sequence and the theoretical residual sequence; The dynamic time warping algorithm is used to calculate the optimal waveform matching path between the actual residual sequence and the theoretical residual sequence in time series. The Euclidean distance or cosine similarity between two sequences is calculated based on the best matching path, and the calculation results are normalized to obtain the morphological similarity.

[0024] This embodiment further restricts the computational logic of the coupled judgment module, employing a dynamic time warping algorithm to address the time lag problem of pathological reactions; and applies the actual residual sequence. Compared with the theoretical residual sequence Perform timeline alignment; specifically, this alignment process includes two steps: First, resampling alignment is achieved by using cubic spline interpolation to adjust the sampling rate of the theoretical residual sequence to match that of the actual residual sequence. Second, amplitude normalization is performed. Z-score standardization is applied to both sequences, subtracting the mean and dividing by the standard deviation to eliminate interference from absolute amplitude differences in waveform morphology assessment. In this step, to prevent division-to-zero errors or noise from being amplified infinitely due to the sequences being in a stable baseline state (i.e., with extremely small standard deviation), the system presets a unique standard deviation safety threshold. If the calculated standard deviation of the sequence If the division operation is skipped, the normalization result of the sequence is directly set to a vector of all zeros. The distance matrix is ​​constructed using a dynamic time warping algorithm, and the cumulative cost is calculated using a recursive formula: in, The boundary condition is set to infinity, and a backtracking method is used to find the optimal matching path from the start to the end of the sequence. This path aims to minimize the regularization cost. The formula is as follows: in, Derived from path planning results, its physical meaning is the length of the matching path; that is, the number of matching point pairs, which usually satisfies... ; The set of points representing the best matching path; : Derived from matrix index, its physical meaning is the first digit on the path. There are 10 matching point pairs, of which... The range of values ​​is , represents the sequence index of the path point; , represents the actual sequence index. With theoretical sequence index The correspondence; The calculation is derived from the square of the Euclidean distance, and the specific formula is as follows: in, Indicates the first path One point pair, and These are the actual and theoretical residual values ​​corresponding to each point pair, respectively, with the physical meaning being the distance metric between the two corresponding points. The normalized distance between the two sequences is calculated based on the optimal matching path and then converted into morphological similarity. The formula is: in, To minimize the cost of regularization; In the asynchronous physiological signal analysis scenario, this embodiment calculates morphological similarity using the DTW algorithm, effectively tolerating the stretching and distortion of real-world vital signs data over time. Even if the drug onset time deviates from the theoretical value due to individual differences, the system can still accurately identify its waveform pattern, ensuring the accuracy of interpretation under time dimension uncertainty.

[0025] Example 6: When determining the anomalous nature of time-series vital sign data, the coupling analysis module is specifically used for: A preset set of judgment thresholds, which includes risk confirmation thresholds; Compare morphological similarity with risk confirmation thresholds; If the morphological similarity is greater than or equal to the risk confirmation threshold, the actual residual sequence is determined to have pathological evolution characteristics, the abnormal nature is determined to be an effective health risk, and a graded intervention instruction is triggered. If the morphological similarity is less than the risk confirmation threshold, the actual residual sequence is determined to be an occasional fluctuation, and the anomaly is identified as environmental noise.

[0026] This embodiment further defines the judgment logic of the coupled judgment module; the system presets a set of judgment thresholds, including risk confirmation thresholds derived from ROC curve analysis of historical confirmed case data. ; hereby it is made clear The specific calculation method is as follows: The system constructs an ROC curve based on a pre-labeled set of pathological / non-pathological historical samples and calculates the Youden index at each point on the curve. The formula is: System selection The similarity threshold corresponding to the maximum value is used as the optimal working point. ,For example This method mathematically guarantees that, in the absence of prior bias, the system's comprehensive ability to distinguish between false alarms and missed alarms reaches its optimal level, thereby establishing an objective risk boundary. The calculated morphological similarity and Comparison; response The system determines that the actual residual sequence possesses pathological evolution characteristics, considers the data anomaly to be driven by a real pathological mechanism, identifies the anomaly as a valid health risk, and triggers a tiered intervention instruction; in response to The system determined that the actual residual sequence was an occasional fluctuation. Although the data deviated from the baseline, its morphology did not conform to the known pathological characteristics, and the abnormality was determined to be environmental noise.

[0027] Example 7: The dynamic intervention module is specifically used for: Obtain the pathological simulation state corresponding to the determination of a valid health risk, and identify the type of pathological perturbation function associated with the pathological simulation state; Based on the pathological perturbation function type, the corresponding clinical intervention path is retrieved from the medical knowledge graph; Based on the deviation of time-series vital signs data, the risk level is determined, and corresponding graded intervention instructions are generated in conjunction with the clinical intervention pathway.

[0028] This embodiment further defines the execution logic of the dynamic intervention module; when a valid health risk is determined, the system obtains the best matching pathological simulation state corresponding to that risk. And identify the type of pathological perturbation function associated with this state. Based on the identified type, the corresponding clinical intervention path is retrieved from the medical knowledge graph; Risk level is calculated based on the deviation of time-series vital sign data. And combined with clinical intervention pathways through mapping functions Generate instructions Risk level here The determination logic is as follows: Calculate the mean absolute percentage error (MAPE) between the time-series vital sign data and the ideal physiological baseline within the abnormal segment. The specific calculation formula is: in, This represents the total number of sampling points within the abnormal time period. For a moment The measured time-series vital signs data, For a moment Ideal physiological baseline values; To prevent extremely small positive numbers with a denominator of zero, the value is taken as... ;like ,set up That is, low risk; if ,set up That is, medium risk; if ,set up This indicates high risk; the mapping function here... Defined as rule-based table lookup logic: the system pre-configures... For index key, with An intervention template library for severity parameters; functions Based on the input Retrieve the corresponding pathology intervention template, such as recommending a follow-up examination or immediate medical attention, and based on... The specific parameters in the numerical template, such as the follow-up interval or medication adjustment dosage, are filled in to generate the final output. Text instructions that include specific action guidelines; In a clinical decision support scenario, this embodiment achieves a closed loop from problem identification to explanation of causes to provision of solutions; the system can directly generate targeted suggestions based on the matched specific pathological model, rather than general abnormality prompts, thus improving the feasibility of health interventions.

[0029] Example 8: The archive data aggregation module is specifically used for: Receive raw vital signs signals from multiple heterogeneous monitoring devices; The original vital signs signals are imputed for missing values ​​and smoothed for outliers to generate standardized time-series vital signs data. The time series vital signs data are periodically decomposed to separate the long-term trend component and the seasonal fluctuation component, and the standardized time series vital signs data after noise removal is used as the main input data for subsequent processing.

[0030] This embodiment further defines the data preprocessing logic of the archive data aggregation module; the system receives raw vital sign signals from multiple heterogeneous monitoring devices such as smart bracelets and home blood pressure monitors; cubic spline interpolation is used to fill in short-term missing data, and sliding window midpoint filtering is used, with a window length of... Based on sampling frequency Dynamically set to cover approximately 0.05 seconds of transient noise, the calculation formula is as follows: in, This indicates a rounding operation, which will cover the theoretical time length of approximately 0.05 seconds of instantaneous noise and the sampling frequency. The floating-point value obtained after multiplication is converted into the nearest discrete integer to determine the specific number of sampling points required for sliding window midpoint filtering, so as to ensure that the window length parameter meets the requirements for integer points in digital signal processing. like The window length is set to 5 sampling points; obvious spike noise is removed to generate standardized time-series vital signs data; the time-series vital signs data are decomposed using STL periodicity. Here, to prevent mathematical logical mutual exclusion, that is, to prevent excessive smoothing of long-term trends from leading to the loss of short-term pathological features, the system imposes key constraints on the parameters of STL decomposition: setting seasonal periodic parameters. Depends on the data sampling frequency of the current processing stage Unit: Hz, Calculation formula is: The system executes a hierarchical resampling strategy here: targeting the long-term trend component. Extraction, pre-sampling the original vital signs signals to That is, 1 point per minute, at this time set To match the circadian rhythm scale; For the waveform morphology analysis involved in the subsequent embodiment 3, the original high-frequency sampling rate is retained. ,like At this point, the corresponding period parameter should be adjusted to For ultra-long period decomposition under such high-frequency sampling, the system preferably adopts parallel computing or sparse sampling estimation based on fast Fourier transform to reduce the computational load. At the same time, set the trend smoothing window parameters. Strictly less than the minimum pathological event duration defined in the medical knowledge graph; for example, if the blood glucose fluctuation cycle caused by drug failure is 3 hours, then set... Corresponding to 2 hours; The length of the trend smoothing window, in units of sampling points, is given by the formula. It is confirmed that, among them, The minimum duration of a pathological event. The sampling frequency; The long-term trend component and the standardized time-series vital signs data after noise removal are output together. The standardized time-series vital signs data serve as the main input data for the subsequent dual-track differential extraction module, while the long-term trend component serves as auxiliary reference data. The separated long-term trend component is configured based on these parameters. In practice, by setting trend smoothing window parameters, it ensures that waveform variation characteristics beyond the expected pathological event timescale are preserved. It only filters out high-frequency noise and fixed 24-hour seasonal fluctuations, while fully preserving the waveform distortion characteristics induced by pathology, thus ensuring the separation of long-term trend components. This reflects the drift of the baseline for chronic diseases; at the same time, standardized time-series vital signs data that retain high-frequency features are directly input into subsequent modules to ensure that there is enough information for morphological similarity calculation.

[0031] Example 9: The system also includes an adaptive correction module, used for: If the abnormality is determined to be environmental noise, the corresponding time-series vital signs data will be marked as non-pathological samples. The parameters for constructing an ideal physiological baseline are fine-tuned using non-pathological samples to update the personalized physiological homeostasis model of the target subjects.

[0032] This embodiment adds an adaptive correction module, giving the system the ability to self-optimize; in response to abnormal properties being determined as environmental noise, the system marks the corresponding time-series vital sign data segments as non-pathological samples; the construction parameters of the ideal physiological benchmark are fine-tuned using non-pathological samples; this embodiment uses a gradient descent method that strictly adheres to dimensional consistency to update the metabolic rate decay variable in the differential equation. The update process is as follows: Initialization: Set the iteration counter. and convergence threshold ; Sensitivity and gradient calculation: based on current parameters Solving the master equation yields The initial conditions for the sensitivity equation are set as follows: This is due to the initial conditions of the master equation. Forced anchoring to Its value is independent of the parameter ,Right now Based on this initial value, solve the sensitivity equation simultaneously: Here we perform a rigorous dimensional analysis: the first term on the right side of the equation Dimensions are The second item middle for Therefore The physical dimensions must be ,Right now rather than simply Calculate the gradient: because Dimensions are ,and Dimensions are Therefore, the gradient term The physical dimensions are Parameter update and dimensional compensation: Execute the update formula: To ensure the second term on the right side of the equation The dimensions of the left side ,dimension Consistency must be eliminated. In And introduce Therefore, the learning rate The dimensions must be 0. This system uses an adaptive normalization formula for calculation. : in, To prevent regularization smoothing factors with zero denominators, the physical quantities are: In this embodiment, a fixed value is used. ; The term is a dimensionless constant, and the denominator is... The dimensions are ; Through this division operation The dimensional calculation process is as follows: This strictly ensures the physical validity of parameter updates and avoids the mathematical error of adding the time dimension to the frequency dimension; Termination judgment: Check If the conditions are met, the updated parameters are output; through this mechanism, the system achieves adaptive evolution of the model while ensuring the correctness of the physical meaning.

[0033] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. An information-based population health record management system, characterized in that: include: The archive data aggregation module is used to collect historical health records of the target object from multiple dimensions, obtain time-series vital sign data and static physiological attribute data, and retrieve preset medical knowledge graphs; The physiological baseline reconstruction module is used to construct the dynamic health trajectory of the target object under ideal and compliant conditions based on static physiological attribute data and medical knowledge graph, which serves as the ideal physiological baseline. The pathological perturbation simulation module is used to define pathological perturbation functions based on medical knowledge graphs and superimpose the pathological perturbation functions onto ideal physiological benchmarks to generate pathological simulation states that include theoretical pathological features. The dual-track differential extraction module is used to calculate the deviation between time-series vital sign data and ideal physiological benchmarks to obtain the actual residual sequence, and to calculate the deviation between the pathological simulation state and ideal physiological benchmarks to obtain the theoretical residual sequence. The coupling analysis module is used to compare the waveform morphology of the actual residual sequence and the theoretical residual sequence, calculate the morphological similarity, and determine the abnormal nature of the time series vital signs data based on the morphological similarity. The dynamic intervention module is used to generate tiered intervention instructions if the abnormality is determined to be a valid health risk, and to perform data filtering operations if the abnormality is determined to be environmental noise.

2. The information-based population health record management system according to claim 1, characterized in that, The physiological baseline reconstruction module is specifically used for: Analyze static physiological attribute data to determine the age and gender parameters of the target object; Based on age and gender parameters, the corresponding natural physiological decline curves are matched from the medical knowledge graph; Based on the natural physiological decline curve and the steady-state rhythm function in the medical knowledge graph, a physiological steady-state differential equation containing the metabolic rate decline variable and the natural physiological drift rate variable is established. Solve the physiological homeostasis differential equation to generate time series curves that do not contain pathological mutation characteristics, which serve as an ideal physiological benchmark.

3. The information-based population health record management system according to claim 1, characterized in that, The pathological disturbance simulation module is specifically used for: Extract pathological causes and violation patterns from medical knowledge graphs and transform them into calculable perturbation factors; A set of pathological perturbation functions is constructed, which includes a concentration drop function for simulating drug adherence failure and a peak decay coefficient for simulating metabolic function deterioration. The pathological perturbation functions in the set of pathological perturbation functions are injected into the ideal physiological benchmark to generate multiple hypothetical pathological curves, and each hypothetical pathological curve is used as a pathological simulation state.

4. The information-based population health record management system according to claim 3, characterized in that, The dual-track differential extraction module is specifically used for: For each pathological simulation state, a difference operation with the ideal physiological benchmark is performed to generate a pure pathological deviation waveform; The morphological features of the pure pathological deviation waveform are extracted, the influence of baseline drift is removed, and the processed waveform data is used as the theoretical residual sequence, which characterizes the theoretical deviation morphology caused by a specific pathological cause.

5. The information-based population health record management system according to claim 1, characterized in that, The coupling analysis module is specifically used for: Time axis alignment is performed between the actual residual sequence and the theoretical residual sequence; The dynamic time warping algorithm is used to calculate the optimal waveform matching path between the actual residual sequence and the theoretical residual sequence in time series. The Euclidean distance or cosine similarity between two sequences is calculated based on the best matching path, and the calculation results are normalized to obtain the morphological similarity.

6. The information-based population health record management system according to claim 5, characterized in that, The coupling analysis module, when determining the anomalous nature of time-series vital sign data, is specifically used for: A preset set of judgment thresholds, which includes risk confirmation thresholds; Compare morphological similarity with risk confirmation thresholds; If the morphological similarity is greater than or equal to the risk confirmation threshold, the actual residual sequence is determined to have pathological evolution characteristics, the abnormal nature is determined to be an effective health risk, and a graded intervention instruction is triggered. If the morphological similarity is less than the risk confirmation threshold, the actual residual sequence is determined to be an occasional fluctuation, and the anomaly is identified as environmental noise.

7. The information-based population health record management system according to claim 6, characterized in that, The dynamic intervention module is specifically used for: Obtain the pathological simulation state corresponding to the determination of a valid health risk, and identify the type of pathological perturbation function associated with the pathological simulation state; Based on the pathological perturbation function type, the corresponding clinical intervention path is retrieved from the medical knowledge graph; Based on the deviation of time-series vital signs data, the risk level is determined, and corresponding graded intervention instructions are generated in conjunction with the clinical intervention pathway.

8. The information-based population health record management system according to claim 1, characterized in that, The archive data aggregation module is specifically used for: Receive raw vital signs signals from multiple heterogeneous monitoring devices; The original vital signs signals are imputed for missing values ​​and smoothed for outliers to generate standardized time-series vital signs data. The time series vital signs data are periodically decomposed to separate the long-term trend component and the seasonal fluctuation component, and the standardized time series vital signs data after noise removal is used as the main input data for subsequent processing.

9. The information-based population health record management system according to claim 1, characterized in that, The system also includes an adaptive correction module for: If the abnormality is determined to be environmental noise, the corresponding time-series vital signs data will be marked as non-pathological samples. The parameters for constructing an ideal physiological baseline are fine-tuned using non-pathological samples to update the personalized physiological homeostasis model of the target subjects.

Images