Classification and Interpretation of ECG Arrhythmia through Deep Learning Techniques

Digitizing healthcare systems is highly essential to identify diseases at an early stage and thus preventing any catastrophes regarding people’s health. One such critical health issue that requires attention in its initial stages itself is Arrhythmia. The research carried out in this paper mainly focuses on Arrhythmia classification and proposes a model to classify ECG signals into different classes of Arrhythmia based on the AAMI standard along with model interpretation. The benchmark ECG MIT-BIH Arrhythmia dataset has been used for training and testing purposes throughout the research. This research proposes the use of various deep learning and data sampling techniques like CNN, RNN, oversampling and under sampling methods in order to build the model that is to be used for Arrhythmia classification. This proposed CNN model outperforms the existing models in terms of different metrics like accuracy, precision, recall, f1-score and reduced prediction time of a sample from 6.23 seconds to 2.09 seconds.