Search

CN-122004887-A - Emergency electrocardio grading early warning method and system based on edge intelligence

CN122004887ACN 122004887 ACN122004887 ACN 122004887ACN-122004887-A

Abstract

The invention relates to an emergency electrocardio grading early warning method and system based on edge intelligence, wherein the method comprises the steps of S1, preprocessing and feature extraction of emergency electrocardio data acquired on an emergency ambulance through wavelet transformation, setting a dynamic threshold value, S2, carrying out anomaly detection and grading early warning on the preprocessed ECG signal through a support vector machine algorithm, triggering multi-stage emergency response according to an early warning level, S3, carrying out fine classification on the heart rhythm signal through a one-dimensional convolutional neural network, identifying various arrhythmia types including normal, supraventricular premature contraction, ventricular premature beat, fusion wave and unknown types, and S4, carrying out fusion decision on multidimensional electrocardio assessment indexes through an intelligent contract system based on a block chain, so as to generate auditable comprehensive risk scores. The invention can realize real-time processing, abnormal detection, fine classification and credible decision on the electrocardiographic data at the edge side of emergency ambulances and the like, reduce the dependence on stable cloud connection and improve the early warning timeliness and reliability.

Inventors

  • WANG SHUE
  • SONG JIANYU
  • CHEN GUANHONG
  • HUANG JI

Assignees

  • 福州大学附属省立医院

Dates

Publication Date
20260512
Application Date
20260131

Claims (10)

  1. 1. An emergency electrocardio grading early warning method based on edge intelligence is characterized by comprising the following steps: Step S1, performing ECG signal preprocessing and feature extraction on emergency electrocardiogram data acquired on an emergency ambulance through wavelet transformation, and completing dynamic threshold setting; s2, carrying out anomaly detection and grading early warning on the preprocessed real-time ECG signal through a support vector machine algorithm, and triggering multi-stage emergency response according to the early warning level; Step 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 and unknown types; And S4, performing fusion decision on the multidimensional electrocardio assessment indexes through an intelligent contract system based on a blockchain to generate auditable comprehensive risk scores.
  2. 2. The emergency electrocardio grading early warning method based on edge intelligence according to claim 1, wherein the step S1 specifically comprises the following steps: s11, preprocessing an original ECG signal by adopting a multiple digital filtering technology, and eliminating motion artifact and baseline drift by deploying a band-pass filter bank of 0.1-200Hz and applying a moving average window to obtain a high-fidelity ECG signal; Step S12, carrying out multi-scale wavelet decomposition on the preprocessed ECG signal, adopting a daubechies-4 wavelet basis to complete 3-level discrete transformation, and reserving a cA3 approximation coefficient and a cD3 detail coefficient to optimize bandwidth and characteristic characterization; And step S13, setting an adaptive dynamic threshold based on the lead and waveform characteristics, wherein the R wave threshold is 1mV, the P wave threshold is 0.08mV and the T wave threshold is 0.1mV, so as to accurately detect the key electrocardiographic waveform.
  3. 3. The emergency electrocardio grading early warning method based on edge intelligence according to claim 1, wherein the step S2 specifically comprises the following steps: S21, constructing a time sequence feature vector based on time domain and morphology key parameters for classification, wherein the key parameters comprise heart rate calculated by R-R interval, quantized P wave amplitude and T wave slope; S22, carrying out real-time classification decision on the time sequence feature vector by configuring a linear support vector machine classifier, and judging abnormal heart beats according to rules of exceeding thresholds in 3 continuous cardiac cycles; Step S23, triggering a multi-level emergency response strategy according to the classification result and a preset priority, starting a local audible and visual alarm for the priority 1 of HR >120bpm, and synchronizing GPS positioning of a patient to an emergency center for the priority 3 of ventricular premature beat, thereby realizing a closed-loop emergency notification of local and cloud cooperation.
  4. 4. The emergency electrocardio grading early warning method based on edge intelligence according to claim 1, wherein the step S3 specifically comprises the following steps: s31, constructing a one-dimensional convolutional neural network 1D-CNN, taking 300 multiplied by 1 dimension wavelet coefficients as input, extracting space-time characteristics through a convolutional layer comprising 32 5 multiplied by 1 convolutional kernels, and sequentially processing through an activation function ReLU, a batch normalization layer and a maximum pooling layer to finish high-dimension characterization of arrhythmia signals; Step S32, based on the extracted high-dimensional representation, outputting probability distribution of five arrhythmia types, namely normal N, supraventricular extra-systole SVEB, ventricular premature beat VEB, fusion wave F and unknown Q through a Softmax activation function; Step S33, based on the fine classification result of step S32, performing dynamic electrocardio characteristic analysis, and quantifying the ST-segment slope change rate and the T-waveform state variation index to enhance the morphological dynamic evaluation and monitoring capability of complex arrhythmia.
  5. 5. The edge intelligence-based emergency electrocardiograph grading early warning method according to claim 4, wherein in the step S31, the dimension of the input matrix of the one-dimensional convolutional neural network is m×n, where M represents the considered time window length, and N represents the number of electrocardiographic channels, the dimension of the convolutional kernel of the one-dimensional convolutional neural network is q×n, where Q represents the time window size covered by the filter, and the convolution operation output size R has a calculation formula: Wherein S represents the step size, i.e. the number of steps the filter moves each time; The activation function ReLU is used to convert the weighted sum of input data into a nonlinear output in the convolutional layer, defined as: Wherein x represents input data of the activation function ReLU; Then, the internal covariate offset is reduced by normalizing the normalized input data in batches, the data expressed as: Wherein, the 、 Mu and sigma 2 respectively represent the mean and variance of the batch, epsilon is a minimum constant avoiding the denominator being 0, gamma and beta are leachable scaling and offset parameters, and the input size and the calculated amount are reduced by maximum pooling, so that the maximum neuron in each activation graph is reserved.
  6. 6. The edge-intelligence-based emergency electrocardiograph grading early warning method according to claim 4, wherein in step S32, the output of the ith neuron is given by assuming that the output layer has K neurons The Softmax activation function is expressed as: , Wherein, the Indicating that the i-th output is subjected to an exponential operation, Representing the summation of index values of all outputs as a normalization factor; the value of the ith output in the probability distribution is represented, the range is (0, 1), and the sum of the softmax values of all outputs is 1.
  7. 7. The emergency electrocardio grading early warning method based on edge intelligence according to claim 1, wherein the step S4 specifically comprises the following steps: s41, constructing a multidimensional electrocardio evaluation function integrating static and dynamic characteristics, and realizing comprehensive risk evaluation by distributing specific weights for heart rate, P wave amplitude, ST-segment variation and T wave slope; Step S42, intelligent contracts are deployed on a private Ethernet chain, multidimensional electrocardio characteristic parameters are stored in an encryption mode, and a linear regression model is embedded to automatically execute risk assessment decision based on the assessment function; and step S43, deploying an audit traceback mechanism, and constructing an untampered audit traceback chain by solidifying and uplink the hash value of the whole-flow data and the first-aid decision log, so that the whole-flow transparency and verifiability of the evaluation and decision process are realized.
  8. 8. The edge intelligence-based emergency electrocardiographic grading early warning method according to claim 1, wherein weights of heart rate, P-wave amplitude, ST-segment variation and T-wave slope are 40%, 30%, 20% and 10%, respectively; the linear regression model is: y = 0.4x 1 + 0.3x 2 + 0.2x 3 + 0.1x 4 Wherein y is a comprehensive risk score, and x 1 、x 2 、x 3 、x 4 corresponds to the heart rate, the P-wave amplitude, the ST-segment variation and the evaluation value of the T-wave slope respectively.
  9. 9. An edge intelligence based emergency electrocardiograph grading early warning system, comprising a memory, a processor and computer program instructions stored on the memory and executable by the processor, when executing the computer program instructions, being capable of implementing the method of any one of claims 1-8.
  10. 10. A computer readable storage medium having stored thereon computer program instructions, which when executed by a processor, implement the method of any of claims 1-8.

Description

Emergency electrocardio grading early warning method and system based on edge intelligence Technical Field The invention relates to the technical fields of medical Internet of things, emergency medical science and artificial intelligence intersection, in particular to an emergency electrocardio grading early warning method and system based on edge intelligence. Background Along with the rapid development of the medical internet of things and artificial intelligence technology, the combination of intelligent perception, edge calculation and real-time diagnosis technology has become an important direction for improving the emergency treatment efficiency. However, in traditional emergency scenarios, there are still some limitations to existing electrocardiographic monitoring and analysis techniques. First, the existing monitoring system highly depends on a cloud center for data processing, the cloud dependency is usually more than 90%, real-time response delay is caused, and the timeliness requirement of critical emergency scenes such as myocardial infarction, malignant arrhythmia and the like on millisecond early warning is difficult to meet. Secondly, most of electrocardiographic analysis models adopt a single classification architecture, and the fine granularity recognition capability of complex arrhythmia is insufficient, so that the failure report rate of critical events is often higher than 5%. Thirdly, the emergency site often faces the problems of unstable network environment, limited bandwidth and the like, so that the interruption rate of key physiological data transmission exceeds 30%, and monitoring continuity and decision reliability are seriously affected. Therefore, how to realize intelligent electrocardio real-time grading early warning with high reliability, low delay and audit under the environment of limited first aid edges of a network becomes a technical problem to be solved urgently. Disclosure of Invention The invention aims to overcome the defects of the prior art, provides an emergency electrocardio grading early warning method and system based on edge intelligence, and the invention can be arranged on the edge side of emergency ambulances and the like, real-time processing, anomaly detection, fine classification and credible decision on electrocardiogram data are realized, dependency on stable cloud connection is obviously reduced, and early warning timeliness and reliability are improved. In order to achieve the purpose, the invention adopts the following technical scheme that the emergency electrocardio grading early warning method based on edge intelligence comprises the following steps: Step S1, performing ECG signal preprocessing and feature extraction on emergency electrocardiogram data acquired on an emergency ambulance through wavelet transformation, and completing dynamic threshold setting; s2, carrying out anomaly detection and grading early warning on the preprocessed real-time ECG signal through a support vector machine algorithm, and triggering multi-stage emergency response according to the early warning level; Step 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 and unknown types; And S4, performing fusion decision on the multidimensional electrocardio assessment indexes through an intelligent contract system based on a blockchain to generate auditable comprehensive risk scores. Further, the step S1 specifically includes the following steps: s11, preprocessing an original ECG signal by adopting a multiple digital filtering technology, and eliminating motion artifact and baseline drift by deploying a band-pass filter bank of 0.1-200Hz and applying a moving average window to obtain a high-fidelity ECG signal; Step S12, carrying out multi-scale wavelet decomposition on the preprocessed ECG signal, adopting a daubechies-4 wavelet basis to complete 3-level discrete transformation, and reserving a cA3 approximation coefficient and a cD3 detail coefficient to optimize bandwidth and characteristic characterization; And step S13, setting an adaptive dynamic threshold based on the lead and waveform characteristics, wherein the R wave threshold is 1mV, the P wave threshold is 0.08mV and the T wave threshold is 0.1mV, so as to accurately detect the key electrocardiographic waveform. Further, the step S2 specifically includes the following steps: S21, constructing a time sequence feature vector based on time domain and morphology key parameters for classification, wherein the key parameters comprise heart rate calculated by R-R interval, quantized P wave amplitude and T wave slope; S22, carrying out real-time classification decision on the time sequence feature vector by configuring a linear support vector machine classifier, and judging abnormal heart beats according to rules of exceedin