ECG Signal Analysis Using 2-D Image Classification with Convolutional Neural Network

Abstract

Cardiovascular diseases, listed as the underlying cause of death, accounted for approximately 836,546 deaths in the United States in 2018. Statistics show that almost one of every three deaths in the US is a result of heart disease. Nearly 2,300 Americans die of cardiovascular disease each day, an average of one death every 38 seconds. This is while quick and immediate action at the onset of such heart conditions can save many lives. To this end, ample research has been reported in the literature on Electrocardiogram (ECG) signal analysis to determine arrhythmia and other cardiac conditions. However, more accurate and near real-time techniques are still under investigation. This work introduces a classifier that will detect abnormalities of the ECG signal with its analysis as a 2-D image fed to a Convolutional Neural Network (CNN) classifier. The proposed method classifies the ECG signal as normal or abnormal by transforming the single-lead ECG signal into images and then applying CNN classification. Images are taken from the European ST-T dataset on PhysioNet databank. Our method yields an accuracy of 97.47%.