Classification of ECG Signals Encrypted with CNN Based Autoencoder with LSTM

The electrocardiogram (ECG) plays a critical role in the early prevention and diagnosis of cardiovascular diseases. The signals obtained from the electrocardiogram device used in the diagnosis of heart diseases are obtained by healthcare professionals and used for arrhythmia detection. Many cardiac abnormalities show up on the ECG, including arrhythmia, which means abnormal heart rhythm. The basis of arrhythmia diagnosis is the identification and correct classification of normal and abnormal heartbeats based on ECG morphology. Autoencoder can compress ECG signals to obtain time series signal features, which can be used for classification purposes. This work is realized using a layered network architecture that includes a convolutional autoencoder for ECG signal compression and a long short-term memory (LSTM) that drives the encoder output for classification. In this architecture, ECG signals are used after preprocessing to achieve higher arrhythmia classification accuracy. The open access MitBih arrhythmia dataset from PhysioNet was used in the study. The results showed that high accuracy, precision and recall were achieved. Compared to traditional methods, this method has better performance in arrhythmia classification.