Fig 1.
Main procedure involved in the classification of ECG.
Table 1.
MIT-BIH verses AAMI 5 heartbeat classes grouping.
Fig 2.
Distribution of heartbeats in different classes of the MIT-BIH ECG database.
Table 2.
Total heartbeats in training dataset classes before and after SMOTE.
Fig 3.
CNN with sequence of layers from input to output.
Fig 4.
(a) Plain network and (b) A residual network.
Fig 5.
The architecture of the proposed ResNet model.
Fig 6.
Training and testing accuracy (batch size = 32).
Fig 7.
Training and testing loss.
Fig 8.
Accuracy using 10-Fold cross validation.
Fig 9.
Classifier performance using confusion matrix (a) Without normalization (b) With normalization.
Fig 10.
Precision and sensitivity values for five classes.
Table 3.
Statistical performance on ECG test dataset.
Table 4.
Performance on ECG test dataset using different batch size for a learning rate of 0.0001.
Table 5.
Performance on ECG test dataset using different batch size for a learning rate of 0.001.
Table 6.
Comparison results with the state of the art.