ECG heartbeat classification using convolutional neural networks and wavelet transform

Abstract

An electrocardiogram is a simple test that can be used to check heart's rhythm and electrical activity and diagnose several abnormal arrhythmias as well. Most of studies try to categorize some sequence of beats and in most successful models the key feature for classification is RR-interval. Our research aims to check whether it is possible to successfully classify ECG heartbeats using scalograms and machine learning algorithms, convolutional neural networks, in particular. All records of necessary signals were taken from open-source PhysioBank Databases from research resource for complex physiologic signals known as PhysioNet. ECG recordings were parsed into sequences of single beats. Due to preprocessing and described model architecture a 92% accuracy has been achieved. Proposed model is still lacking some performance in comparison with state-of-the-art solutions in ECG heart categorization. However, it is possible to modify applied approach.