An improved cardiac arrhythmia classification using stationary wavelet transform decomposed short duration QRS segment and Bi-LSTM network

Arrhythmia is a kind of cardiac conduction disorder those result in irregular heartbeats. The electrocardiograph (ECG) signal may identify conduction system abnormalities. However, its visual analysis is challenging and time-consuming. An automated system for cardiac disorder detection may help in early and prompt diagnosis of diseases. In this paper, stationary wavelet transform (SWT) was used for pre-processing of the raw ECG signal before the segmentation and normalization process. Thereafter, recurrent neural network (RNN), gated recurrent units (GRU), bi-directional long short-term memory (Bi-LSTM) have been implemented for classification of normal, left bundle branch block (L-BBB), right bundle branch block l(R-BBB), premature atrial contraction (PAC), and premature ventricular contraction (PVC) beats. Bi-LSTM networks have shown best accuracy of 99.72% among all three implemented models. This demonstrates that this model is appropriate for computer-aided diagnosis of heartbeats.

Automated summary

Arrhythmia is a cardiac conduction disorder that leads to irregular heartbeats. The visual analysis of electrocardiograph (ECG) signals is challenging and time-consuming, but an automated system can aid in the early and prompt diagnosis of diseases. This paper used the stationary wavelet transform (SWT) for pre-processing the raw ECG signals and implemented recurrent neural networks (RNN), gated recurrent units (GRU), and bi-directional long short-term memory (Bi-LSTM) for beat classification. Among the three models, the Bi-LSTM network achieved the highest accuracy of 99.72%, indicating its suitability for computer-aided diagnosis of heartbeats.