Emergency electrocardiogram grading early warning method and system based on edge intelligence

Abstract

The invention relates to an emergency treatment electrocardiogram grading early warning method and system based on edge intelligence, and the method comprises the steps: S1, carrying out the preprocessing and feature extraction of emergency treatment electrocardiogram data collected on an emergency ambulance through wavelet transform, and setting a dynamic threshold value; s2, anomaly detection and graded early warning are carried out on the preprocessed ECG signals through a support vector machine algorithm, and multi-stage emergency response is triggered according to early warning levels; s3, finely classifying the heart rhythm signals through a one-dimensional convolutional neural network, and identifying a plurality of arrhythmia types, including normal, supraventricular premature contraction, ventricular premature beat, fusion waves and unknown types; and S4, performing fusion decision on the multi-dimensional electrocardiogram evaluation indexes through an intelligent contract system based on a block chain, and generating an auditable comprehensive risk score. According to the method, real-time processing, anomaly detection, fine classification and credible decision-making of electrocardiogram data can be realized on the edge side of an emergency ambulance and the like, the dependence on stable connection of a cloud end is reduced, and the early warning timeliness and reliability are improved.

Description

Technical Field

[0001] This invention relates to the fields of medical Internet of Things, emergency medicine and artificial intelligence, specifically to an emergency electrocardiogram triage and early warning method and system based on edge intelligence. Background Technology

[0002] With the rapid development of medical IoT and AI technologies, combining intelligent sensing, edge computing, and real-time diagnostic technologies has become an important direction for improving the efficiency of emergency treatment. However, in traditional emergency scenarios, existing ECG monitoring and analysis technologies still have some limitations.

[0003] First, existing monitoring systems heavily rely on cloud-based data centers for processing, with cloud dependency typically exceeding 90%. This leads to real-time response delays, making it difficult to meet the millisecond-level early warning requirements of critical care scenarios such as myocardial infarction and malignant arrhythmias. Second, most electrocardiogram (ECG) analysis models employ a single classification architecture, lacking the ability to fine-grainedly identify complex arrhythmias, resulting in a false negative rate for critical events often exceeding 5%. Third, emergency sites frequently face issues such as unstable network environments and limited bandwidth, causing a critical physiological data transmission interruption rate exceeding 30%, severely impacting monitoring continuity and decision-making reliability.

[0004] Therefore, how to achieve highly reliable, low-latency, and auditable intelligent real-time tiered early warning of electrocardiograms in emergency edge environments with limited network coverage has become an urgent technical problem to be solved. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide an emergency electrocardiogram (ECG) triage and early warning method and system based on edge intelligence. This invention can realize real-time processing, anomaly detection, fine classification and reliable decision-making of ECG data at the edge, such as in ambulances, significantly reducing the dependence on stable cloud connections and improving the timeliness and reliability of early warning.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: an emergency electrocardiogram triage and early warning method based on edge intelligence, comprising the following steps:

[0007] Step S1: Perform ECG signal preprocessing and feature extraction on the emergency ECG data collected in the ambulance using wavelet transform, and complete the dynamic threshold setting;

[0008] Step S2: The preprocessed real-time ECG signal is subjected to anomaly detection and graded early warning using the support vector machine algorithm, and a multi-level emergency response is triggered according to the early warning level;

[0009] Step S3: Finely classify cardiac rhythm signals using a one-dimensional convolutional neural network to identify various types of arrhythmias, including normal, supraventricular premature contractions, ventricular premature beats, fusion waves, and unknown types;

[0010] Step S4: Use a blockchain-based smart contract system to integrate and make decisions on multi-dimensional ECG assessment indicators to generate an auditable comprehensive risk score.

[0011] Furthermore, step S1 specifically includes the following steps:

[0012] Step S11: The original ECG signal is preprocessed using multiple digital filtering techniques. By deploying a 0.1-200Hz bandpass filter bank and applying a moving average window, motion artifacts and baseline drift are eliminated to obtain a high-fidelity ECG signal.

[0013] Step S12: Perform multi-scale wavelet decomposition on the preprocessed ECG signal, use the daubechies-4 wavelet basis to complete the 3-level discrete transformation, and retain the cA3 approximation coefficients and cD3 detail coefficients to optimize bandwidth and feature representation.

[0014] Step S13: Set adaptive dynamic thresholds based on lead and waveform characteristics, where the R wave threshold is 1mV, the P wave threshold is 0.08mV, and the T wave threshold is 0.1mV, in order to accurately detect key ECG waveforms.

[0015] Furthermore, step S2 specifically includes the following steps:

[0016] Step S21: Construct a temporal feature vector based on key temporal and morphological parameters for classification. The key parameters include heart rate calculated from the RR interval, quantized P wave amplitude, and T wave slope.

[0017] Step S22: Configure a linear support vector machine classifier to perform real-time binary classification decision on the temporal feature vector, and determine abnormal heartbeats based on the rule that three consecutive cardiac cycles exceed the threshold.

[0018] Step S23: Trigger a multi-level emergency response strategy based on the classification results and preset priorities. For priority 1 cases with HR>120bpm, activate the local audible and visual alarm. For priority 3 cases of premature ventricular contractions, synchronize the patient's GPS location to the emergency center to achieve closed-loop emergency notification between local and cloud.

[0019] Furthermore, step S3 specifically includes the following steps:

[0020] Step S31: Construct a one-dimensional convolutional neural network 1D-CNN, using 300×1-dimensional wavelet coefficients as input, extracting spatiotemporal features through a convolutional layer containing 32 5×1 convolutional kernels, and then processing them sequentially through the activation function ReLU, batch normalization layer and max pooling layer to complete the high-dimensional representation of arrhythmia signals.

[0021] Step S32: Based on the extracted high-dimensional representation, output the probability distribution of five arrhythmia types, namely normal N, supraventricular premature contraction SVEB, ventricular premature beat VEB, fusion wave F, and unknown Q, through the Softmax activation function;

[0022] Step S33: Based on the detailed classification results in step S32, perform dynamic electrocardiogram feature analysis to quantify the ST segment slope change rate and T wave morphology variation index, so as to enhance the ability to dynamically assess and monitor complex arrhythmias.

[0023] Further, in step S31, the input matrix of the one-dimensional convolutional neural network has a dimension of M×N, where M represents the length of the considered time window and N represents the number of ECG channels; the convolution kernel of the one-dimensional convolutional neural network has a dimension of Q×N, where Q represents the size of the time-domain window covered by the filter; the output size R of the convolution operation is calculated using the following formula:

[0024]

[0025] Where S represents the step size, that is, the number of steps the filter moves each time;

[0026] The ReLU activation function is used in convolutional layers to transform a weighted sum of input data into a non-linear output, and is defined as follows:

[0027]

[0028] Where x represents the input data for the ReLU activation function;

[0029] Then, by batch normalizing the input data to reduce internal covariate bias, the data is represented as follows:

[0030]

[0031] in, , μ and σ represent the data before and after batch normalization, respectively. 2 Let represent the mean and variance of the batch, respectively; ε is a minimal constant to avoid a denominator of 0; and γ and β are learnable scaling and offset parameters. Max pooling is used to reduce the input size and computational cost, while retaining the neurons with the maximum value in each activation graph.

[0032] Furthermore, in step S32, assuming the output layer has K neurons, the output of the i-th neuron is... The Softmax activation function is then expressed as:

[0033] ,

[0034] in, This indicates that the exponentiation operation is performed on the i-th output. This represents the summation of the exponential values ​​of all outputs, used as a normalization factor; Let represent the value of the i-th output in the probability distribution, ranging from (0,1), and the sum of the softmax values ​​of all outputs is 1.

[0035] Furthermore, step S4 specifically includes the following steps:

[0036] Step S41: Construct a multi-dimensional ECG assessment function that integrates static and dynamic features. By assigning specific weights to heart rate, P wave amplitude, ST segment variability, and T wave slope, a comprehensive risk assessment can be achieved.

[0037] Step S42: Deploy a smart contract on a private Ethereum chain to store multi-dimensional ECG feature parameters in an encrypted manner, and embed a linear regression model to automatically execute risk assessment decisions based on the evaluation function;

[0038] Step S43: Deploy an audit traceability mechanism. By solidifying the hash values ​​of the entire process data and the emergency response decision logs onto the blockchain, an immutable audit traceability chain is constructed, achieving full transparency and verifiability of the assessment and decision-making process.

[0039] Furthermore, the weights of heart rate, P wave amplitude, ST segment variability, and T wave slope are 40%, 30%, 20%, and 10%, respectively.

[0040] The linear regression model is as follows:

[0041] y = 0.4x1 + 0.3x2 + 0.2x3 + 0.1x4

[0042] Where y is the comprehensive risk score, and x1, x2, x3, and x4 are the assessment values ​​of heart rate, P wave amplitude, ST segment variability, and T wave slope, respectively.

[0043] The present invention also provides an emergency electrocardiogram triage and early warning system based on edge intelligence, including a memory, a processor, and computer program instructions stored in the memory and capable of being executed by the processor. When the processor executes the computer program instructions, it can implement the above-mentioned method.

[0044] The present invention also provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the above-described method.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] 1. Low latency and high reliability: By deploying core algorithms such as wavelet transform, SVM fast early warning, and 1D-CNN fine classification on the edge of the ambulance, this invention significantly reduces the dependence on data uploading to the cloud, realizes millisecond-level local real-time analysis and response, and effectively overcomes the risk of interruption caused by network instability.

[0047] 2. Graded and precise early warning: This invention adopts a two-level machine learning architecture that combines SVM rapid screening with 1D-CNN fine classification. First, it achieves rapid capture and graded early warning of abnormal heartbeats with low computational overhead. Then, it performs fine-grained identification of complex arrhythmias in accordance with medical standards, thus balancing speed and accuracy.

[0048] 3. Credible and auditable decision-making: This invention innovatively introduces blockchain smart contract technology to solidify the assessment model, decision-making logic, and key data hash values ​​on the chain, ensuring the transparency, immutability, and full traceability of the risk assessment process, and providing credible technical support for emergency response decision-making.

[0049] 4. Lightweight and Scalable System: The 1D-CNN model in this invention is designed for one-dimensional time-series signals, has a small number of parameters, and is suitable for deployment on edge devices. Simultaneously, the system reserves a federated learning interface, supporting model optimization and personalized training while protecting privacy, and possesses good scalability. Attached Figure Description

[0050] Figure 1 This is a flowchart illustrating the implementation of the emergency electrocardiogram triage and early warning method based on edge intelligence provided in this embodiment of the invention. Detailed Implementation

[0051] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0052] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0053] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0054] like Figure 1 As shown in the figure, this embodiment provides an emergency electrocardiogram triage and early warning method based on edge intelligence, and its specific implementation steps are as follows.

[0055] Step S1: Perform ECG signal preprocessing and feature extraction on the emergency ECG data collected in the ambulance using wavelet transform, and complete the dynamic threshold setting. The implementation steps are as follows.

[0056] Step S11: The original ECG signal is preprocessed using multiple digital filtering techniques. By deploying a 0.1-200Hz bandpass filter bank and applying a moving average window, motion artifacts and baseline drift are effectively eliminated to obtain a high-fidelity ECG signal.

[0057] Step S12: Perform multi-scale wavelet decomposition on the preprocessed ECG signal, use the daubechies-4 wavelet basis to complete the 3-level discrete transformation, and retain the cA3 approximation coefficients and cD3 detail coefficients to optimize bandwidth and feature representation.

[0058] Step S13: Set adaptive dynamic thresholds based on lead and waveform characteristics, where the R wave threshold (Lead I) is 1mV and the P / T wave thresholds (Lead II) are 0.08mV and 0.1mV respectively, in order to accurately detect key ECG waveforms.

[0059] Step S2: Anomaly detection and graded early warning are performed on the preprocessed real-time ECG signal using the Support Vector Machine (SVM) algorithm, and multi-level emergency response is triggered according to the early warning level. The implementation steps are as follows.

[0060] Step S21: Construct a temporal feature vector based on key temporal and morphological parameters for classification. The key parameters include heart rate calculated from the RR interval, quantized P wave amplitude (0.08±0.02mV), and T wave slope.

[0061] Step S22: Configure a linear support vector machine (SVM) classifier (C=1.0, kernel=linear) to perform real-time binary classification decision on the time-series feature vectors, and determine abnormal heartbeats based on the rule that three consecutive cardiac cycles exceed the threshold.

[0062] Step S23: Trigger a multi-level emergency response strategy based on the classification results and preset priorities. For priority 1 (HR>120bpm), activate the local audible and visual alarm. For priority 3 (ventricular premature beats), synchronize the patient's GPS location to the emergency center to achieve closed-loop emergency notification between local and cloud.

[0063] Step S3: Fine-classify the cardiac rhythm signal using a one-dimensional convolutional neural network (1D-CNN) to identify various types of arrhythmias, including normal (N), supraventricular premature contractions (SVEB), ventricular premature beats (VEB), fusion beats (F), and unknown (Q) types. The implementation steps are as follows.

[0064] Step S31: Construct a one-dimensional convolutional neural network (1D-CNN), using 300×1-dimensional wavelet coefficients as input, extracting spatiotemporal features through a convolutional layer containing 32 5×1 convolutional kernels, and then processing them sequentially through the activation function ReLU, batch normalization layer and max pooling layer to complete the high-dimensional representation of arrhythmia signals.

[0065] The input matrix of the one-dimensional convolutional neural network has dimensions M×N, where M represents the length of the time window considered (M=300) and N represents the number of ECG channels (N=1). The convolution kernel of the one-dimensional convolutional neural network has dimensions Q×N, where Q represents the size of the time window covered by the filter (Q=5). The output size R of the convolution operation is calculated using the following formula:

[0066]

[0067] Where S represents the step size (i.e., the number of steps the filter moves each time).

[0068] The ReLU activation function is used in convolutional layers to transform a weighted sum of input data into a non-linear output, and is defined as follows:

[0069]

[0070] Where x represents the input data for the ReLU activation function.

[0071] Then, the input data is standardized by batch normalization (BN) to reduce internal covariate bias, and the data is represented as follows:

[0072]

[0073] in, , μ and σ represent the data before and after batch normalization, respectively. 2 Let denot mean and variance of the batch, respectively, and ε be a minimal constant to avoid a denominator of 0 (e.g., 10).−5 ), γ and β are learnable scaling and offset parameters; Max Pooling is used to reduce input size and computational cost while retaining neurons with the highest values ​​in each activation graph.

[0074] Step S32: Based on the extracted high-dimensional representation, output the probability distribution of five arrhythmia types: normal (N), supraventricular premature contractions (SVEB), ventricular premature beats (VEB), fusion beats (F), and unknown (Q) through the Softmax activation function.

[0075] Assuming the output layer has K neurons, the output of the i-th neuron is The Softmax activation function can then be expressed as:

[0076] ,

[0077] in, This indicates that the exponentiation operation is performed on the i-th output. This represents the summation of the exponential values ​​of all outputs, used as a normalization factor; Let represent the value of the i-th output in the probability distribution, ranging from (0,1), and the sum of the softmax values ​​of all outputs is 1.

[0078] Step S33: Based on the detailed classification results in step S32, perform dynamic electrocardiogram feature analysis to quantify the ST segment slope change rate and T wave morphology variation index, so as to enhance the ability to dynamically assess and monitor complex arrhythmias.

[0079] Step S4: A blockchain-based smart contract system is used to integrate multi-dimensional ECG assessment indicators to generate an auditable comprehensive risk score. The implementation steps are as follows.

[0080] Step S41: Construct a multi-dimensional ECG assessment function that integrates static and dynamic features. By assigning specific weights to heart rate, P wave amplitude, ST segment variability, and T wave slope, a comprehensive risk assessment can be achieved.

[0081] In this embodiment, the weights of heart rate, P wave amplitude, ST segment variability, and T wave slope are set to 40%, 30%, 20%, and 10%, respectively.

[0082] Step S42: Deploy a smart contract on a private Ethereum chain to store multi-dimensional ECG feature parameters in an encrypted manner, and embed a linear regression model to automatically execute risk assessment decisions based on the evaluation function.

[0083] The linear regression model is expressed as:

[0084] y = 0.4x1 + 0.3x2 + 0.2x3 + 0.1x4

[0085] Where y is the comprehensive risk score, and x1, x2, x3, and x4 are the assessment values ​​of heart rate, P wave amplitude, ST segment variability, and T wave slope, respectively.

[0086] Step S43: Deploy an audit traceability mechanism. By solidifying the hash values ​​of the entire process data and the emergency response decision logs onto the blockchain, an immutable audit traceability chain is constructed, achieving full transparency and verifiability of the assessment and decision-making process.

[0087] This embodiment also provides an emergency electrocardiogram triage and early warning system based on edge intelligence, including a memory, a processor, and computer program instructions stored in the memory and executable by the processor. When the processor executes the computer program instructions, it can implement the above-described method.

[0088] This embodiment also provides a computer-readable storage medium storing computer program instructions that, when executed by a processor, implement the above-described method.

[0089] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0090] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0091] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1The function specified in one or more boxes.

[0092] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0093] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. An emergency electrocardiogram (ECG) triage and early warning method based on edge intelligence, characterized in that, Includes the following steps: Step S1: Perform ECG signal preprocessing and feature extraction on the emergency ECG data collected in the ambulance using wavelet transform, and complete the dynamic threshold setting; Step S2: The preprocessed real-time ECG signal is subjected to anomaly detection and graded early warning using the support vector machine algorithm, and a multi-level emergency response is triggered according to the early warning level; Step S3: Finely classify cardiac rhythm signals using a one-dimensional convolutional neural network to identify various types of arrhythmias, including normal, supraventricular premature contractions, ventricular premature beats, fusion waves, and unknown types; Step S4: Use a blockchain-based smart contract system to integrate and make decisions on multi-dimensional ECG assessment indicators to generate an auditable comprehensive risk score.

2. The emergency electrocardiogram triage and early warning method based on edge intelligence according to claim 1, characterized in that, Step S1 specifically includes the following steps: Step S11: The original ECG signal is preprocessed using multiple digital filtering techniques. By deploying a 0.1-200Hz bandpass filter bank and applying a moving average window, motion artifacts and baseline drift are eliminated to obtain a high-fidelity ECG signal. Step S12: Perform multi-scale wavelet decomposition on the preprocessed ECG signal, use the daubechies-4 wavelet basis to complete the 3-level discrete transformation, and retain the cA3 approximation coefficients and cD3 detail coefficients to optimize bandwidth and feature representation. Step S13: Set adaptive dynamic thresholds based on lead and waveform characteristics, where the R wave threshold is 1mV, the P wave threshold is 0.08mV, and the T wave threshold is 0.1mV, in order to accurately detect key ECG waveforms.

3. The emergency electrocardiogram triage and early warning method based on edge intelligence according to claim 1, characterized in that, Step S2 specifically includes the following steps: Step S21: Construct a temporal feature vector based on key temporal and morphological parameters for classification. The key parameters include heart rate calculated from the RR interval, quantized P wave amplitude, and T wave slope. Step S22: Configure a linear support vector machine classifier to perform real-time binary classification decision on the temporal feature vector, and determine abnormal heartbeats based on the rule that three consecutive cardiac cycles exceed the threshold. Step S23: Trigger a multi-level emergency response strategy based on the classification results and preset priorities. For priority 1 cases with HR>120bpm, activate the local audible and visual alarm. For priority 3 cases of premature ventricular contractions, synchronize the patient's GPS location to the emergency center to achieve closed-loop emergency notification between local and cloud.

4. The emergency electrocardiogram triage and early warning method based on edge intelligence according to claim 1, characterized in that, Step S3 specifically includes the following steps: Step S31: Construct a one-dimensional convolutional neural network 1D-CNN, using 300×1-dimensional wavelet coefficients as input, extracting spatiotemporal features through a convolutional layer containing 32 5×1 convolutional kernels, and then processing them sequentially through the activation function ReLU, batch normalization layer and max pooling layer to complete the high-dimensional representation of arrhythmia signals. Step S32: Based on the extracted high-dimensional representation, output the probability distribution of five arrhythmia types, namely normal N, supraventricular premature contraction SVEB, ventricular premature beat VEB, fusion wave F, and unknown Q, through the Softmax activation function; Step S33: Based on the detailed classification results in step S32, perform dynamic electrocardiogram feature analysis to quantify the ST segment slope change rate and T wave morphology variation index, so as to enhance the ability to dynamically assess and monitor complex arrhythmias.

5. The emergency electrocardiogram triage and early warning method based on edge intelligence according to claim 4, characterized in that, In step S31, the input matrix of the one-dimensional convolutional neural network has a dimension of M×N, where M represents the length of the considered time window and N represents the number of ECG channels; the convolution kernel of the one-dimensional convolutional neural network has a dimension of Q×N, where Q represents the size of the time-domain window covered by the filter; the output size R of the convolution operation is calculated using the following formula: Where S represents the step size, that is, the number of steps the filter moves each time; The ReLU activation function is used in convolutional layers to transform a weighted sum of input data into a non-linear output, and is defined as follows: Where x represents the input data for the ReLU activation function; Then, by batch normalizing the input data to reduce internal covariate bias, the data is represented as follows: in, , μ and σ represent the data before and after batch normalization, respectively. 2 Let represent the mean and variance of the batch, respectively; ε is a minimal constant to avoid a denominator of 0; and γ and β are learnable scaling and offset parameters. Max pooling is used to reduce the input size and computational cost, while retaining the neurons with the maximum value in each activation graph.

6. The emergency electrocardiogram triage and early warning method based on edge intelligence according to claim 4, characterized in that, In step S32, assuming the output layer has K neurons, the output of the i-th neuron is... The Softmax activation function is then expressed as: ， in, This indicates that the exponentiation operation is performed on the i-th output. This represents the summation of the exponential values ​​of all outputs, used as a normalization factor; Let represent the value of the i-th output in the probability distribution, ranging from (0,1), and the sum of the softmax values ​​of all outputs is 1.

7. The emergency electrocardiogram triage and early warning method based on edge intelligence according to claim 1, characterized in that, Step S4 specifically includes the following steps: Step S41: Construct a multi-dimensional ECG assessment function that integrates static and dynamic features. By assigning specific weights to heart rate, P wave amplitude, ST segment variability, and T wave slope, a comprehensive risk assessment can be achieved. Step S42: Deploy a smart contract on a private Ethereum chain to store multi-dimensional ECG feature parameters in an encrypted manner, and embed a linear regression model to automatically execute risk assessment decisions based on the evaluation function; Step S43: Deploy an audit traceability mechanism. By solidifying the hash values ​​of the entire process data and the emergency response decision logs onto the blockchain, an immutable audit traceability chain is constructed, achieving full transparency and verifiability of the assessment and decision-making process.

8. The emergency electrocardiogram triage and early warning method based on edge intelligence according to claim 1, characterized in that, The weights for heart rate, P wave amplitude, ST segment variability, and T wave slope are 40%, 30%, 20%, and 10%, respectively. The linear regression model is as follows: y = 0.4x1 + 0.3x2 + 0.2x3 + 0.1x4 Where y is the comprehensive risk score, and x1, x2, x3, and x4 are the assessment values ​​of heart rate, P wave amplitude, ST segment variability, and T wave slope, respectively.

9. An emergency electrocardiogram triage and early warning system based on edge intelligence, characterized in that, It includes a memory, a processor, and computer program instructions stored in the memory and executable by the processor, wherein when the processor executes the computer program instructions, it can implement the method as described in any one of claims 1-8.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by a processor, the method described in any one of claims 1-8 is implemented.

Images