Sudden death first-aid monitoring and early-warning method, system and equipment based on multi-source data and medium

Abstract

The invention relates to the technical field of first-aid monitoring and early-warning, in particular to a sudden death first-aid monitoring and early-warning method, system and equipment based on multi-source data and a medium. The method comprises the steps that baseline data of a patient with coronary heart disease, real-time data of electrocardiogram monitoring equipment, first-aid dispatching data, hospital diagnosis and treatment data and vital sign monitoring data are collected; integrating the multi-source data through a standardized interface and a data cleaning algorithm to form a unified database; analyzing the integrated data, and predicting the risk value of the sudden death risk of the coronary heart disease patient by constructing a sudden death risk prediction model; screening out middle and high risk groups based on the predicted risk value, and sending early warning information to patients, family members and community medical institutions when the risk value exceeds a set threshold value; middle-high risk groups are screened out in time, early warning is triggered, first-aid time is won for patients, and the sudden death occurrence rate is reduced.

Description

Technical Field

[0001] This application relates to the field of emergency medical monitoring and early warning technology, specifically to a method, system, device, and medium for emergency medical monitoring and early warning of sudden death based on multi-source data. Background Technology

[0002] With the accelerating pace of modern life and the exacerbation of population aging, cardiovascular diseases, especially coronary heart disease, have become one of the major threats to human health. Sudden death, as one of the most serious complications of coronary heart disease, is characterized by its suddenness and high mortality rate, placing a heavy burden on patients' families and society. Traditional methods of warning and emergency treatment for sudden death often rely on doctors' experience and patients' self-perception, lacking real-time and accurate data support, making it difficult to intervene effectively before sudden death occurs.

[0003] In recent years, with the rapid development of information technology and medical equipment, the acquisition and integration of multi-source data has become possible. This data includes baseline data of coronary heart disease patients, real-time data from electrocardiogram monitoring devices, emergency dispatch data, hospital treatment data, and vital sign monitoring data. Through in-depth mining and analysis of this multi-source data, the risk of sudden death in coronary heart disease patients can be predicted more accurately, thus providing a scientific basis for timely and effective emergency measures.

[0004] However, there is currently a lack of a sudden death emergency monitoring and early warning system based on multi-source data that can achieve real-time monitoring, risk assessment, and early warning issuance for patients with coronary heart disease. Summary of the Invention

[0005] In view of the lack of efficient integration in the existing emergency medical system, which makes it difficult to achieve early warning of sudden death risk and thus carry out rapid and accurate rescue, this invention provides a method, system, device and medium for monitoring and early warning of sudden death based on multi-source data.

[0006] In a first aspect, the present invention provides a method for monitoring and early warning of sudden cardiac arrest based on multi-source data, comprising the following steps: We collect baseline data from coronary heart disease patients, real-time data from electrocardiogram monitoring devices, emergency dispatch data, hospital diagnosis and treatment data, and vital sign monitoring data. We then integrate the multi-source data through standardized interfaces and data cleaning algorithms to form a unified database. The integrated data was analyzed, and a sudden death risk prediction model was constructed to predict the risk value of sudden death in patients with coronary heart disease. Based on the predicted risk values, medium- and high-risk groups are screened out, and warning information is sent to patients, their families, and community medical institutions when the risk values ​​exceed the set threshold.

[0007] By integrating data from multiple sources to form a unified database, comprehensive data support is provided for subsequent risk prediction and early warning. This method can fully utilize information from various data sources, improving the accuracy and reliability of sudden death risk prediction.

[0008] As a further limitation of the technical solution of the present invention, the step of analyzing the integrated data and predicting the risk value of sudden death in patients with coronary heart disease by constructing a sudden death risk prediction model includes: Features related to sudden death risk are extracted from multi-source data, and significant features are screened through univariate analysis. The contribution of feature combination is evaluated by multivariate modeling, and the features are then dimensionality reduced. Select an algorithm model, train the model based on the training set, and optimize the model hyperparameters through cross-validation; The patient's real-time data is input into the trained model, and combined with baseline data and dynamic features, the risk value of patients with coronary heart disease is predicted.

[0009] By constructing a sudden death risk prediction model, it is possible to quantitatively assess the risk of sudden death in patients with coronary heart disease. This method can combine baseline and real-time data of patients to dynamically adjust the risk prediction results, providing a scientific basis for early warning issuance and emergency response measures.

[0010] As a further limitation of the technical solution of this invention, the steps of extracting features related to sudden death risk from multi-source data, screening significant features through univariate analysis, evaluating the contribution of feature combinations through multivariate modeling, and performing dimensionality reduction on the features include: Multi-source data is divided into patient baseline data, dynamic physiological data, clinical diagnosis and treatment data, and emergency event data to identify potential characteristics related to sudden death risk in each category; Numerical and categorical features are directly extracted from patient baseline data and clinical diagnosis and treatment data, and their derived statistics are calculated. Time-domain, frequency-domain, and nonlinear features were extracted from ECG monitoring data. Trend statistics were calculated for vital signs signals using a sliding window, and outliers were detected using the isolated forest algorithm. Keywords were extracted from electronic medical records using the BERT model. A knowledge graph was constructed based on co-occurrence frequency and semantic similarity. Subgraphs related to sudden death were searched in the knowledge graph to quantify the strength of the association. Significant features are screened through univariate analysis, and the contribution of feature combinations is evaluated by multivariate modeling. The features are then subjected to dimensionality reduction.

[0011] Extracting features related to sudden death risk from multi-source data and then filtering and reducing them through univariate analysis and multivariate modeling can remove redundant information and improve the model's predictive performance. This method ensures that the model focuses only on the features most relevant to sudden death risk, thereby improving the accuracy and efficiency of predictions.

[0012] As a further limitation of the technical solution of this invention, the steps of dimensionality reduction of features by screening significant features through univariate analysis and evaluating the contribution of feature combinations through multivariate modeling include: Each feature and the sudden death outcome are independently statistically tested, and significant features with p-values ​​less than a preset threshold are selected to generate a significant feature set; The contribution of feature combinations is evaluated by a multivariate model, and features with low contribution or a set contribution threshold are removed from the significant feature set to generate a significant feature subset. For the selected significant feature subset, principal component analysis is used to project the original features into a low-dimensional space to obtain the dimensionality-reduced feature matrix.

[0013] By combining univariate analysis and multivariate modeling, significant features can be more accurately identified and a subset of significant features can be generated. This approach ensures that the model uses only features that significantly contribute to the prediction of sudden death risk, further improving the accuracy and reliability of the prediction.

[0014] As a further limitation of the technical solution of this invention, in the step of evaluating the contribution of feature combinations through a multivariate model, eliminating features with low contribution or a set contribution threshold from the significant feature set, and generating a significant feature subset, the multivariate model is a random forest model, specifically including: The significant features selected from the univariate analysis are standardized, the categorical features are encoded, and a feature matrix suitable for input to the random forest model is constructed. Initialize the random forest model parameters and train the model. During model training, record the number of splits and the reduction of the Gini index for each feature in each decision tree. Calculate the average Gini index reduction for each feature across all decision trees to obtain the contribution of each feature; The contribution of all features is normalized, a contribution threshold is set, and features with a contribution value lower than the contribution threshold are removed from the significant feature set to generate a significant feature subset.

[0015] Using a random forest model to evaluate the contribution of feature combinations can more accurately quantify the importance of each feature in predicting sudden death risk. This method can eliminate features with low contribution, reduce model complexity, and improve prediction efficiency and accuracy.

[0016] As a further limitation of the technical solution of the present invention, the step of projecting the original features into a low-dimensional space using principal component analysis to obtain the dimensionality-reduced feature matrix after screening the significant feature subset includes: The features in the selected significant feature subset are organized into a feature matrix; Standardize the feature matrix; Calculate the covariance matrix; Eigenvalue decomposition of the covariance matrix yields eigenvalues ​​and corresponding eigenvectors; Calculate the cumulative variance contribution rate of the first k principal components; Select the first k eigenvectors to form the projection matrix; The original feature matrix is ​​projected onto a new low-dimensional space based on the projection matrix to obtain the dimensionality-reduced feature matrix.

[0017] Principal component analysis (PCA) is used to reduce the dimensionality of the selected salient feature subset, which preserves the main information in the original data while reducing the data dimensionality. This method can reduce the computational complexity of the model, improve the efficiency and accuracy of prediction, and also facilitate data visualization and interpretation.

[0018] As a further limitation of the technical solution of the present invention, the steps of screening out medium- and high-risk groups based on predicted risk values ​​and sending early warning information to patients, their families, and community medical institutions when the risk values ​​exceed a set threshold include: Based on preset risk thresholds, the population is divided into three categories: low-risk, medium-risk, and high-risk. Real-time warnings are triggered for medium- and high-risk groups, and warning information is sent to patients, their families, and medical institutions via mobile applications, SMS, or system push notifications.

[0019] Based on preset risk thresholds, the population is divided into low-risk, medium-risk, and high-risk categories, and real-time alerts are triggered for medium- and high-risk groups, enabling precise management and timely intervention for patients with coronary heart disease. This method can send alert information to patients, their families, and medical institutions via mobile applications, SMS, or system push notifications, improving the speed and efficiency of emergency response.

[0020] Secondly, the present invention also provides a sudden cardiac arrest monitoring and early warning system based on multi-source data, comprising: The multi-source data acquisition and integration module is used to collect baseline data of coronary heart disease patients, real-time data from electrocardiogram monitoring equipment, emergency dispatch data, hospital diagnosis and treatment data, and vital sign monitoring data. It integrates multi-source data through standardized interfaces and data cleaning algorithms to form a unified database. The high-risk population risk assessment module is used to analyze the integrated data and predict the risk value of coronary heart disease patients by building a sudden death risk prediction model; The risk warning module is used to screen out medium- and high-risk groups based on predicted risk values ​​and send warning information to patients, their families and community medical institutions when the risk value exceeds a set threshold. Thirdly, the present invention also provides an electronic device, the electronic device comprising: at least one processor; and a memory communicatively connected to the at least one processor; the memory storing computer program instructions executable by the at least one processor, the computer program instructions being executed by the at least one processor to enable the at least one processor to execute the sudden death emergency monitoring and early warning method based on multi-source data as described in the first aspect.

[0021] Fourthly, the present invention also provides a non-transitory computer-readable storage medium that stores computer instructions that cause the computer to execute the sudden death emergency monitoring and early warning method based on multi-source data as described in the first aspect.

[0022] As can be seen from the above technical solutions, this application has the following advantages: By integrating multi-source data (coronary artery disease baseline data, electrocardiogram monitoring data, emergency dispatch data, etc.), a sudden death risk prediction model is constructed, significantly improving the sensitivity and specificity of risk prediction. Based on real-time data and dynamic characteristics, high-risk groups are promptly screened and early warnings are triggered, buying valuable emergency time for patients and reducing the incidence of sudden death. Attached Figure Description To more clearly illustrate the technical solution of this application, the accompanying drawings used in the description will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a flowchart illustrating the method provided in an embodiment of the present invention.

[0024] Figure 2 This is a system block diagram provided for an embodiment of the present invention. Detailed Implementation

[0025] To make the purpose, features, and advantages of this application more apparent and understandable, specific embodiments and accompanying drawings will be used to clearly and completely describe the technical solution protected by this application. Obviously, the embodiments described below are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0026] like Figure 1As shown in the figure, this invention provides a method for monitoring and early warning of sudden cardiac arrest based on multi-source data, comprising the following steps: S1. Collect baseline data of patients with coronary heart disease, real-time data from electrocardiogram monitoring equipment, emergency dispatch data, hospital diagnosis and treatment data, and vital sign monitoring data, and integrate the multi-source data through standardized interfaces and data cleaning algorithms to form a unified database; Data is collected in real time from Hospital Information System (HIS), Electronic Medical Record (EMR), Electrocardiogram (ECG) monitoring equipment, Emergency Dispatch System (120), and wearable devices via standardized interfaces (such as HL7 and FHIR). Data cleaning algorithms are used to process missing values ​​(multiple imputation), outliers (3σ principle), and duplicate records (patient unique identifier deduplication), and all data is converted to a unified timestamp format (UTC time).

[0027] S2. Analyze the integrated data and predict the risk value of coronary heart disease patients by constructing a sudden death risk prediction model; In this step, numerical features such as age, gender, and BMI, as well as categorical features such as family medical history and history of myocardial infarction, are extracted from baseline data. Temporal / frequency domain features such as QT interval and heart rate variability (HRV) are extracted from electrocardiogram (ECG) data. Significant features with p-values ​​<0.05 are screened using chi-square tests (for categorical variables) and t-tests (for continuous variables). Random forest is used to evaluate the contribution of feature combinations, and features with contribution values ​​below a threshold are removed. Principal component analysis (PCA) is performed on the screened feature matrix, retaining principal components with a cumulative variance contribution rate ≥95%.

[0028] S3. Based on the predicted risk value, identify medium- and high-risk groups and send early warning information to patients, their families and community medical institutions when the risk value exceeds the set threshold. The model inputs real-time patient data and outputs a risk score (0-1). A threshold is set (e.g., ≥0.7 for medium to high risk), triggering an alert and sending it to the patient and medical institution via mobile application and SMS.

[0029] In this embodiment of the invention, the steps of analyzing the integrated data and predicting the risk value of coronary heart disease patients by constructing a sudden death risk prediction model include: S21. Extract features related to sudden death risk from multi-source data, screen significant features through univariate analysis, evaluate the contribution of feature combinations using multivariate modeling, and perform dimensionality reduction on the features; this step specifically includes: S211. Divide multi-source data into patient baseline data, dynamic physiological data, clinical diagnosis and treatment data, and emergency event data, and identify potential characteristics related to sudden death risk in each category; (1) Patient baseline data Numerical data: age, gender, BMI, blood pressure, blood sugar, cholesterol, etc.

[0030] Classification: Family medical history (yes / no), history of myocardial infarction (yes / no), smoking status (yes / no), etc.

[0031] (2) Dynamic physiological data ECG monitoring: heart rate, QT interval, ST segment elevation / depression, heart rate variability (HRV) indicators (such as SDNN, RMSSD).

[0032] Vital signs: respiratory rate, blood oxygen saturation, body temperature, etc.

[0033] (3) Clinical diagnosis and treatment data Examination results: abnormal electrocardiogram markers, myocardial enzyme levels, and coronary angiography results.

[0034] Medication records: Use of drugs such as beta-blockers and aspirin.

[0035] (4) Emergency event data 120 dispatch record: call time, location, preliminary diagnosis (e.g., chest pain, difficulty breathing).

[0036] Previous emergency events: frequency of attacks, emergency response time, and whether an AED was used.

[0037] S212. Directly extract numerical and categorical features from patient baseline data and clinical diagnosis and treatment data, and calculate their derived statistics; here, derived statistics refer to variance and mean. S213. Extract time-domain, frequency-domain, and nonlinear features from ECG monitoring data; calculate trend statistics for vital signs signals using a sliding window; and use the isolated forest algorithm to detect outliers. S214. Use the BERT model to extract keywords from electronic medical records, construct a knowledge graph based on co-occurrence frequency and semantic similarity, search for subgraphs related to sudden death in the knowledge graph, and quantify the association strength; use a pre-trained BERT model to encode the text. Extract medical-related terms (such as "coronary heart disease," "myocardial infarction," "chest pain," and "aspirin") using Named Entity Recognition (NER) technology. Output: A structured list of keywords (such as [coronary heart disease, chest pain, aspirin, ventricular tachycardia]).

[0038] The purpose of constructing a knowledge graph based on co-occurrence frequency and semantic similarity is to model extracted keywords through entity relationships, forming a knowledge graph that clarifies the associations between diseases, symptoms, and drugs. Specifically, it involves statistically analyzing the number of times keywords co-occur in the same medical record or patient record (e.g., "coronary heart disease" and "chest pain" often appear together). High-frequency co-occurring keyword pairs (e.g., [coronary heart disease, chest pain]) are used as potential association edges. The semantic similarity of keywords is calculated using a BERT model (e.g., comparing word vectors using cosine similarity). For example, although "myocardial infarction" and "angina pectoris" do not co-occur directly, they are highly semantically related and can be associated. Nodes: Disease (coronary heart disease), Symptom (chest pain), Drug (aspirin), Examination indicator (QT interval prolongation). Edges: Relationships defined based on co-occurrence frequency and semantic similarity (e.g., "coronary heart disease → chest pain," "aspirin → coronary heart disease"). Weights: The weights of the edges are determined by the weighted sum of co-occurrence frequency and semantic similarity.

[0039] The purpose of quantifying risk associations through subgraph matching is to search for subgraphs in a knowledge graph that are related to a target risk (such as sudden death) and quantify the strength of the association. Graph Neural Networks (GNNs) or similarity-based matching algorithms are used to find paths in the knowledge graph that are highly similar to the subgraph patterns. The weights of the matching subgraphs (e.g., the product of edge weights) are calculated as a quantitative indicator of the risk association strength. High-risk subgraphs and their association scores (e.g., a score of 0.89 for [coronary artery disease + chest pain + QT interval prolongation] indicates a high risk of sudden death) are defined.

[0040] S215. Use univariate analysis to screen significant features, combine multivariate modeling to evaluate the contribution of feature combinations, and perform dimensionality reduction on the features.

[0041] Steps S212-S214 above are all feature engineering steps, and S215 is to process and reduce the dimensionality of the features obtained in the above steps.

[0042] S22. Select an algorithm model, train the model based on the training set, and optimize the model hyperparameters through cross-validation; The model outputs a risk probability between 0 and 1, maximizing the model's accuracy in predicting the risk of sudden death, while also taking into account interpretability and computational efficiency.

[0043] loss function

[0044] in For the number of categories, For the sample size, For the first Each sample belongs to category The true distribution probability, To predict probabilities. This application uses three categories: high, medium, and low risk.

[0045] S23. Input the patient's real-time data into the trained model, and combine it with baseline data and dynamic features to predict the risk value of patients with coronary heart disease.

[0046] In some embodiments, the steps of dimensionality reduction of features include screening significant features through univariate analysis, evaluating the contribution of feature combinations through multivariate modeling, and performing dimensionality reduction on the features. S2151. Perform independent statistical tests on each feature and the sudden death outcome, and select significant features with p-values ​​less than a preset threshold to generate a significant feature set. The p-value is a probabilistic indicator used in statistics to measure the degree of inconsistency between observed data and the null hypothesis. Its core function is to help researchers determine whether the experimental results are statistically significant.

[0047] S2152. Evaluate the contribution of feature combinations through a multivariate model, and remove features with low contribution from the set of significant features to generate a subset of significant features; S2153. For the selected significant feature subset, principal component analysis is used to project the original features into a low-dimensional space to obtain the dimensionality-reduced feature matrix.

[0048] In some embodiments, in the step of evaluating the contribution of feature combinations using a multivariate model and generating a subset of significant features by removing features with low contribution from the set of significant features, the multivariate model is a random forest model, specifically including: The significant features selected from the univariate analysis are standardized, the categorical features are encoded, and a feature matrix suitable for input to the random forest model is constructed. Initialize the random forest model parameters and train the model. During model training, record the number of splits and the reduction of the Gini index for each feature in each decision tree. Calculate the average Gini index reduction for each feature across all decision trees to obtain the contribution of each feature; where the feature... Contribution

[0049] in, For the number of trees, Features In the The first tree The reduction in the Gini index caused by each split node; The contribution of all features is normalized, a contribution threshold is set, and features with contributions below the threshold are removed from the salient feature set to generate a salient feature subset. Normalization is achieved by converting the contribution to the [0,1] interval and dividing each contribution by the maximum contribution value.

[0050] In some embodiments, the step of projecting the original features into a low-dimensional space using principal component analysis to obtain the dimensionality-reduced feature matrix after filtering the salient feature subset includes: The features in the selected significant feature subset are organized into a feature matrix. Each row represents a sample (such as a patient), and each column represents a feature (such as age, heart rate, blood pressure, etc.). For the characteristic matrix Perform standardization processing; for each feature Calculate its mean and standard deviation Then, the value of each sample on that feature. Convert to ; Calculate the covariance matrix, where the i-th element in the covariance matrix is... Line number Column elements Representation of features With features The covariance is calculated using the following formula: ,in It is the first Each sample in features Standardized values ​​on It is a feature The standardized mean.

[0051] Eigenvalue decomposition of the covariance matrix yields eigenvalues ​​and their corresponding eigenvectors; eigenvalues The eigenvectors represent the variance explained by each principal component. Indicates the direction of the principal component; Before calculation Cumulative variance contribution rate of each principal component Among them, the cumulative variance contribution rate reflects the previous The proportion of total variance explained by each principal component; based on a preset cumulative variance contribution rate threshold (usually 90%, 95%, or 99%, etc.), determine the number of principal components to retain. For example, if we choose a cumulative variance contribution rate of 95%, then we find the minimum one. Make ≥0.95 Before selection eigenvectors constitute a projection matrix Each column is a feature vector.

[0052] The original feature matrix Projecting onto the new low-dimensional space yields the dimensionality-reduced feature matrix. ,in, ,in It is A matrix, where each row represents a sample in Coordinates on each principal component.

[0053] In some embodiments, the steps of screening out medium- and high-risk groups based on predicted risk values ​​and sending early warning information to patients, their families, and community healthcare institutions when the risk values ​​exceed a set threshold include: Based on preset risk thresholds, the population is divided into three categories: low-risk, medium-risk, and high-risk. Real-time warnings are triggered for medium- and high-risk groups, and warning information is sent to patients, their families, and medical institutions via mobile applications, SMS, or system push notifications.

[0054] like Figure 2 As shown, this embodiment of the invention also provides a sudden cardiac arrest monitoring and early warning system based on multi-source data, comprising: The multi-source data acquisition and integration module is used to collect baseline data of coronary heart disease patients, real-time data from electrocardiogram monitoring equipment, emergency dispatch data, hospital diagnosis and treatment data, and vital sign monitoring data. It integrates multi-source data through standardized interfaces and data cleaning algorithms to form a unified database. The high-risk population risk assessment module is used to analyze the integrated data and predict the risk value of coronary heart disease patients by building a sudden death risk prediction model; The risk warning module is used to screen out medium- and high-risk groups based on predicted risk values ​​and send warning information to patients, their families and community medical institutions when the risk value exceeds a set threshold. The visual data management module is used to display risk assessments and early warning information.

[0055] In some embodiments, the high-risk population risk assessment module includes a feature processing unit, a model building unit, and a risk prediction unit; The feature processing unit is used to extract features related to the risk of sudden death from multi-source data, screen significant features through univariate analysis, evaluate the contribution of feature combination by combining multivariate modeling, and perform dimensionality reduction on the features. The model building unit selects an algorithm model, trains the model based on the training set, and optimizes the model hyperparameters through cross-validation. The risk prediction unit is used to input real-time patient data into a trained model, and combine baseline data and dynamic features to predict the risk value of patients with coronary heart disease. In some embodiments, the feature processing unit includes: The first processing submodule divides the multi-source data into patient baseline data, dynamic physiological data, clinical diagnosis and treatment data, and emergency event data, and identifies the potential characteristics related to the risk of sudden death in each category; The second processing submodule directly extracts numerical and categorical features from patient baseline data and clinical diagnosis and treatment data, and calculates their derived statistics. The third processing submodule extracts time-domain, frequency-domain, and nonlinear features from the ECG monitoring data, calculates trend statistics for vital signs signals using a sliding window, and uses the isolated forest algorithm to detect outliers. The fourth processing submodule uses the BERT model to extract keywords from electronic medical records, constructs a knowledge graph based on co-occurrence frequency and semantic similarity, searches for subgraphs related to sudden death in the knowledge graph, and quantifies the strength of the association. The dimensionality reduction module filters significant features through univariate analysis, evaluates the contribution of feature combinations through multivariate modeling, and performs dimensionality reduction on the features.

[0056] In some embodiments, the dimensionality reduction processing module includes: The initial screening subunit is used to perform independent statistical tests on each feature and the sudden death outcome, and to select significant features with p-values ​​less than a preset threshold to generate a significant feature set; The secondary screening subunit is used to evaluate the contribution of feature combinations through a multivariate model and remove features with low contribution from the significant feature set to generate a significant feature subset. The dimensionality reduction subunit is used to project the original features into a low-dimensional space using principal component analysis to obtain the dimensionality-reduced feature matrix from the selected salient feature subset.

[0057] In some embodiments, the multivariate model is a random forest model. The secondary screening subunit is specifically used to standardize the significant features after univariate analysis screening, encode the classification features, and construct a feature matrix suitable for input to the random forest model; initialize the random forest model parameters for model training, and record the number of splits and the reduction of Gini index of each feature in each decision tree during model training; calculate the average reduction of Gini index of each feature in all decision trees to obtain the contribution of each feature; normalize the contribution of all features, set a contribution threshold, and remove features with a contribution lower than the contribution threshold from the significant feature set to generate a significant feature subset.

[0058] In some embodiments, the dimensionality reduction subunit is specifically used to organize the features in the filtered significant feature subset into a feature matrix; standardize the feature matrix; calculate the covariance matrix; perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and corresponding eigenvectors; and perform pre-calculation... Cumulative variance contribution rate of each principal component; before selection The eigenvectors form a projection matrix; the original eigenma matrix is ​​projected onto a new low-dimensional space based on the projection matrix to obtain the dimensionality-reduced eigenma matrix.

[0059] In some embodiments, the risk warning unit is used to divide the population into three categories: low-risk, medium-risk and high-risk according to a preset risk threshold, trigger real-time warnings for medium- and high-risk groups, and send warning information to patients, their families and medical institutions through mobile applications, SMS or system push. This invention also provides an electronic device, comprising: a processor, a communication interface, a memory, and a communication bus, wherein the processor, communication interface, and memory communicate with each other via the communication bus. The communication bus can be used for information transmission between the electronic device and sensors. The processor can call logical instructions in the memory to execute the following methods: S1, collecting baseline data of coronary heart disease patients, real-time data from electrocardiogram monitoring devices, emergency dispatch data, hospital diagnosis and treatment data, and vital sign monitoring data, and integrating the multi-source data through a standardized interface and data cleaning algorithm to form a unified database; S2, analyzing the integrated data and predicting the risk value of coronary heart disease patients by constructing a sudden death risk prediction model; S3, screening out medium- and high-risk groups based on the predicted risk value, and sending early warning information to patients, their families, and community medical institutions when the risk value exceeds a set threshold.

[0060] Furthermore, the logical instructions in the aforementioned memory can be implemented as software functional units and sold or used as independent products, and can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0061] This invention provides a non-transitory computer-readable storage medium storing computer instructions that cause a computer to execute the methods provided in the above-described method embodiments. These instructions include, for example: S1, collecting baseline data of coronary heart disease patients, real-time data from electrocardiogram monitoring devices, emergency dispatch data, hospital treatment data, and vital sign monitoring data, and integrating the multi-source data through standardized interfaces and data cleaning algorithms to form a unified database; S2, analyzing the integrated data and predicting the risk value of coronary heart disease patients by constructing a sudden death risk prediction model; S3, screening out medium- and high-risk groups based on the predicted risk values, and sending warning information to patients, their families, and community medical institutions when the risk value exceeds a set threshold.

[0062] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for monitoring and early warning of sudden cardiac arrest based on multi-source data, characterized in that, Includes the following steps: We collect baseline data from coronary heart disease patients, real-time data from electrocardiogram monitoring devices, emergency dispatch data, hospital diagnosis and treatment data, and vital sign monitoring data. We then integrate the multi-source data through standardized interfaces and data cleaning algorithms to form a unified database. The integrated data was analyzed, and a sudden death risk prediction model was constructed to predict the risk value of sudden death in patients with coronary heart disease. Based on the predicted risk values, medium- and high-risk groups are identified, and warning information is sent to patients, their families, and community medical institutions when the risk values ​​exceed the set threshold.

2. The method for monitoring and early warning of sudden cardiac arrest based on multi-source data according to claim 1, characterized in that, The steps involved in analyzing the integrated data and predicting the risk value of sudden death in patients with coronary heart disease by constructing a sudden death risk prediction model include: Features related to sudden death risk are extracted from multi-source data, and significant features are screened through univariate analysis. The contribution of feature combination is evaluated by multivariate modeling, and the features are then dimensionality reduced. Select an algorithm model, train the model based on the training set, and optimize the model hyperparameters through cross-validation; The patient's real-time data is input into the trained model, which combines baseline data and dynamic features to predict the risk value of patients with coronary heart disease.

3. The method for monitoring and early warning of sudden cardiac arrest based on multi-source data according to claim 2, characterized in that, The steps involved in extracting features related to sudden death risk from multi-source data, screening significant features through univariate analysis, evaluating the contribution of feature combinations through multivariate modeling, and performing dimensionality reduction on the features include: Multi-source data is divided into patient baseline data, dynamic physiological data, clinical diagnosis and treatment data, and emergency event data to identify potential characteristics related to sudden death risk in each category; Numerical and categorical features are directly extracted from patient baseline data and clinical diagnosis and treatment data, and their derived statistics are calculated. Time-domain, frequency-domain, and nonlinear features were extracted from ECG monitoring data. Trend statistics were calculated for vital signs signals using a sliding window, and outliers were detected using the isolated forest algorithm. Keywords were extracted from electronic medical records using the BERT model. A knowledge graph was constructed based on co-occurrence frequency and semantic similarity. Subgraphs related to sudden death were searched in the knowledge graph to quantify the strength of the association. Significant features are screened through univariate analysis, and the contribution of feature combinations is evaluated by multivariate modeling. The features are then subjected to dimensionality reduction.

4. The method for monitoring and early warning of sudden cardiac arrest based on multi-source data according to claim 3, characterized in that, The steps for feature dimensionality reduction include: screening significant features through univariate analysis, evaluating the contribution of feature combinations through multivariate modeling, and then performing feature dimensionality reduction. Each feature and the sudden death outcome are independently statistically tested, and significant features with p-values ​​less than a preset threshold are selected to generate a significant feature set; The contribution of feature combinations is evaluated by a multivariate model, and features with low contribution or a set contribution threshold are removed from the significant feature set to generate a significant feature subset. For the selected significant feature subset, principal component analysis is used to project the original features into a low-dimensional space to obtain the dimensionality-reduced feature matrix.

5. The method for monitoring and early warning of sudden cardiac arrest based on multi-source data according to claim 4, characterized in that, In the step of evaluating the contribution of feature combinations using a multivariate model, removing features with low contribution or those exceeding a set contribution threshold from the salient feature set, and generating a salient feature subset, the multivariate model is a random forest model, specifically including: The significant features selected from the univariate analysis are standardized, the categorical features are encoded, and a feature matrix suitable for input to the random forest model is constructed. Initialize the random forest model parameters and train the model. During model training, record the number of splits and the reduction of the Gini index for each feature in each decision tree. Calculate the average Gini index reduction for each feature across all decision trees to obtain the contribution of each feature; The contribution of all features is normalized, a contribution threshold is set, and features with a contribution value lower than the contribution threshold are removed from the significant feature set to generate a significant feature subset.

6. The method for monitoring and early warning of sudden cardiac arrest based on multi-source data according to claim 5, characterized in that, The steps for projecting the original features into a low-dimensional space using principal component analysis to obtain the dimensionality-reduced feature matrix after filtering the salient feature subset include: The features in the selected significant feature subset are organized into a feature matrix; Standardize the feature matrix; Calculate the covariance matrix; Eigenvalue decomposition of the covariance matrix yields eigenvalues ​​and corresponding eigenvectors; Calculate the cumulative variance contribution rate of the first k principal components; Select the first k eigenvectors to form the projection matrix; The original feature matrix is ​​projected onto a new low-dimensional space based on the projection matrix to obtain the dimensionality-reduced feature matrix.

7. The method for monitoring and early warning of sudden cardiac arrest based on multi-source data according to claim 6, characterized in that, The steps for identifying medium- and high-risk individuals based on predicted risk values ​​and sending early warning information to patients, their families, and community healthcare institutions when the risk value exceeds a set threshold include: Based on preset risk thresholds, the population is divided into three categories: low-risk, medium-risk, and high-risk. Real-time warnings are triggered for medium- and high-risk groups, and warning information is sent to patients, their families, and medical institutions via mobile applications, SMS, or system push notifications.

8. A sudden cardiac arrest monitoring and early warning system based on multi-source data, characterized in that, include: The multi-source data acquisition and integration module is used to collect baseline data of coronary heart disease patients, real-time data from electrocardiogram monitoring equipment, emergency dispatch data, hospital diagnosis and treatment data, and vital sign monitoring data. It integrates multi-source data through standardized interfaces and data cleaning algorithms to form a unified database. The high-risk population risk assessment module is used to analyze the integrated data and predict the risk value of coronary heart disease patients by building a sudden death risk prediction model; The risk warning module is used to screen out medium- and high-risk groups based on predicted risk values ​​and send warning information to patients, their families and community medical institutions when the risk value exceeds a set threshold.

9. An electronic device, characterized in that, The electronic device includes: at least one processor; and a memory communicatively connected to the at least one processor; the memory stores computer program instructions executable by the at least one processor, the computer program instructions being executed by the at least one processor to enable the at least one processor to perform the sudden death emergency monitoring and early warning method based on multi-source data as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium stores computer instructions that cause the computer to execute the sudden death emergency monitoring and early warning method based on multi-source data as described in any one of claims 1 to 7.

Images