Automatic recognition and classification method for electrocardiogram heartbeat based on artificial intelligence
An automatic recognition and classification method for electrocardiogram heartbeat based on artificial intelligence, comprising: processing a received original electrocardiogram digital signal to obtain heartbeat time sequence data and lead heartbeat data; cutting the lead heartbeat data according to the heartbeat time sequence data to generate lead heartbeat analysis data; performing data combination on the lead heartbeat analysis data to obtain a one-dimensional heartbeat analysis array; performing data dimension amplification and conversion according to the one-dimensional heartbeat analysis array to obtain four-dimensional tensor data; and inputting the four-dimensional tensor data to a trained LepuEcgCatNet heartbeat classification model, to obtain heartbeat classification information. The method overcomes the defect that the conventional method only depends on single lead independent analysis for result summary statistics and thus classification errors are more easily obtained, and the accuracy of the electrocardiogram heartbeat classification is greatly improved.
1. An automatic identification and classification method of ECG heart beat based on artificial intelligence, comprising:
processing received original ECG digital signals to generate heart beat time sequence data and lead heart beat data;
cutting the lead heart beat data with a set data amount to generate lead heart beat analysis data according to the heart beat time sequence data;
performing data combination on the lead heart beat analysis data to obtain a one-dimensional heart beat analysis array;
performing data dimension amplification conversion on the one-dimensional heart beat analysis array to obtain four-dimensional tensor data; and
inputting the four-dimensional tensor data into a trained LepuEcgCatNet heart beat classification model to obtain heart beat classification information corresponding to the lead heart beat analysis data.
2. The automatic identification and classification method of ECG heart beat according to claim 1 , wherein the cutting the lead heart beat data with a set data amount to generate lead heart beat analysis data according to the heart beat time sequence data, comprises:
determining a center sample point of the lead heart beat data according to the heart beat time sequence data; and
taking the center sample point as a center, according to time characterization data and a preset data sample frequency, data sampling the lead heart beat data to both sides with the set data amount, so as to obtain the lead heart beat analysis data.
3. The automatic identification and classification method of ECG heart beat according to claim 1 , wherein the lead heart beat analysis data is single-lead heart beat analysis data, and the performing data combination on the lead heart beat analysis data to obtain a one-dimensional heart beat analysis array comprises:
combining the single-lead heart beat analysis data into the one-dimensional heart beat analysis array according to the heart beat time sequence data.
4. The automatic identification and classification method of ECG heart beat according to claim 1 , wherein the lead heart beat analysis data is multi-lead heart beat analysis data, and the performing data combination on the lead heart beat analysis data to obtain a one-dimensional heart beat analysis array comprises:
combining the multi-lead heart beat analysis data into the one-dimensional heart beat analysis array according to lead parameters and the heart beat time sequence data.
5. The automatic identification and classification method of ECG heart beat according to claim 1 , wherein the performing data dimension amplification conversion on the one-dimensional heart beat analysis array to obtain four-dimensional tensor data comprises:
converting the one-dimensional heart beat analysis array into a four-dimensional tensor data input format required by the trained LepuEcgCatNet heart beat classification model in a specific conversion mode; wherein the four-dimensional tensor data has four factors, which are height data, width data, channel data and batch data;
wherein the specific conversion mode comprises:
determining the height data and the width data according to a length of input samples, and generating fusion data according to the height data, the width data or the channel data; wherein the channel data is a number of leads; and
generating the four-dimensional tensor data according to the batch data and the fusion data; wherein the batch data is a number of the input samples.
6. The automatic identification and classification method of ECG heart beat according to claim 1 , wherein the inputting the four-dimensional tensor data into a trained LepuEcgCatNet heart beat classification model to obtain heart beat classification information corresponding to the lead heart beat analysis data comprises:
according to the trained LepuEcgCatNet heart beat classification model, performing layer-by-layer convolution feature extraction on the input four-dimensional tensor data, and obtaining the heart beat classification information corresponding to the lead heart beat analysis data through reasoning operation.
7. The automatic identification and classification method of ECG heart beat according to claim 6 , the method further comprises: establishing and training the trained LepuEcgCatNet heart beat classification model.
8. The automatic identification and classification method of ECG heart beat according to claim 7 , the establishing and training the trained LepuEcgCatNet heart beat classification model comprises:
based on deep convolutional neural networks AlexNet, Vgg16, ResNet, Inception and open mature deep learning models, constructing an end-to-end multi-labeled LepuEcgCatNet heart beat classification model which is a deep learning model having multi-layer convolutional neural network for the automatic identification and classification method of the ECG heart beat; and
training the end-to-end multi-labeled LepuEcgCatNet heart beat classification model to obtain model structure data and parameter data, and storing and encrypting the model structure data and the parameter data.
9. The method according to claim 7 , the training comprises:
selecting training samples;
converting the training samples into a preset standard data format for storage; and
performing data cutting, combining and converting to the training samples, inputting the training samples into the trained LepuEcgCatNet heart beat classification model for training, storing trained model structure data and parameter data according to Google Protocol Buffers data protocol, and encrypting the trained model structure data and the parameter data for protecting by using a symmetric encryption algorithm.
10. The automatic identification and classification method of ECG heart beat according to claim 2 , wherein the lead heart beat analysis data is single-lead heart beat analysis data, and the performing data combination on the lead heart beat analysis data to obtain a one-dimensional heart beat analysis array comprises:
combining the single-lead heart beat analysis data into the one-dimensional heart beat analysis array according to the heart beat time sequence data.
11. The automatic identification and classification method of ECG heart beat according to claim 10 , wherein the lead heart beat analysis data is multi-lead heart beat analysis data, and the performing data combination on the lead heart beat analysis data to obtain a one-dimensional heart beat analysis array comprises:
combining the multi-lead heart beat analysis data into the one-dimensional heart beat analysis array according to lead parameters and the heart beat time sequence data.
12. The automatic identification and classification method of ECG heart beat according to claim 11 , wherein the performing data dimension amplification conversion on the one-dimensional heart beat analysis array to obtain four-dimensional tensor data comprises:
converting the one-dimensional heart beat analysis array into a four-dimensional tensor data input format required by the trained LepuEcgCatNet heart beat classification model in a specific conversion mode; wherein the four-dimensional tensor data has four factors, which are height data, width data, channel data and batch data;
wherein the specific conversion mode comprises:
determining the height data and the width data according to a length of input samples, and generating fusion data according to the height data, the width data or the channel data; wherein the channel data is a number of leads; and
generating the four-dimensional tensor data according to the batch data and the fusion data; wherein the batch data is a number of the input samples.
13. The automatic identification and classification method of ECG heart beat according to claim 12 , wherein the inputting the four-dimensional tensor data into a trained LepuEcgCatNet heart beat classification model to obtain heart beat classification information corresponding to the lead heart beat analysis data comprises:
according to the trained LepuEcgCatNet heart beat classification model, performing layer-by-layer convolution feature extraction on the input four-dimensional tensor data, and obtaining the heart beat classification information corresponding to the lead heart beat analysis data through reasoning operation.
14. The automatic identification and classification method of ECG heart beat according to claim 13 , the method further comprises: establishing and training the trained LepuEcgCatNet heart beat classification model.
15. The automatic identification and classification method of ECG heart beat according to claim 14 , the establishing and training the LepuEcgCatNet heart beat classification model comprises:
based on deep convolutional neural networks AlexNet, Vgg16, ResNet, Inception and open mature deep learning models, constructing an end-to-end multi-labeled LepuEcgCatNet heart beat classification model which is a deep learning model having multi-layer convolutional neural network for the automatic identification and classification method of the ECG heart beat; and
training the end-to-end multi-labeled LepuEcgCatNet heart beat classification model to obtain model structure data and parameter data, and storing and encrypting the model structure data and the parameter data.
16. The method according to claim 14 , the training comprises:
selecting training samples;
converting the training samples into a preset standard data format for storage; and
performing data cutting, combining and converting to the training samples, inputting the training samples into the end-to-end multi-labeled LepuEcgCatNet heart beat classification model for training, storing trained model structure data and parameter data according to Google Protocol Buffers data protocol, and encrypting the trained model structure data and the parameter data for protecting by using a symmetric encryption algorithm.