ECG Heartbeat Arrhythmias Classification Using Convolutional Neural Networks

Abstract

While cardiac diseases are increasing in the past years, heart monitoring has become crucial to assess the heart behavior and detect any arrhythmia if available. Electrocardiograms or ECG signals, are records of the electrical activity of the heart that illustrates the way the depolarization wave flow in each heartbeat; A proper study of an ECG signal’s characteristics is the gold standard of providing effective diagnostics for cardiac diseases. The aim of this work is to provide an automatic approach of analyzing and detecting arrhythmias using deep neural networks or to be more specific, 2-dimensional convolutional neural networks. The great performance in extracting the spatial features of input image data is what contributed in CNNs popularity and is what makes it more suitable than any other model. In this study a CNN architecture was proposed and discussed in terms of performance, and the impact of deep learning techniques which are batch normalization and dropout on it