Novel Methodology for Cardiac Arrhythmias Classification Based on Long-Duration ECG Signal Fragments Analysis

Abstract

According to the reports published by various organizations, it can be seen that about 50 million people are at risk of cardiovascular diseases (CVDs) around the world. Moreover, different types of heart diseases are the most common causes of mortality. This chapter, therefore, investigates a cardiac disorders database (ECG) with 17 classes (normal sinus rhythm, the rhythm of the pacemaker, and fifteen arrhythmias) using a novel classification methodology. The data set is based on long-duration ECG signal fragments. The Electrocardiography (ECG) is a very popular process to record the electrical activity of the heart during specific time. Even though there are a lot of studies in the literature, however, there are many other open issues in the topic. The main objective of the current study is to present a new and efficient methods in order to do automatic recognition of myocardium dysfunctions. The proposed methods are introduced that can be used in different situations such as mobile devices, telemedicine, cloud computing, and finally supporting preventive and supportive treatment of CVDs. Since the performance of proposed algorithms is very important, the time duration, as an additional criterion, is also analyzed in real time. The obtained outcomes demonstrate that our methodology has outstanding performance compared to the methods presented in the literature. This study uses 744 fragments of ECG signal database related to 29 patients from the MIH-BIH Arrhythmia database (only for one lead—MLII). By using Welch’s method and a discrete Fourier transform, the spectral power density is predicted in order to increase the characteristic features of the ECG signals. The research presents a new evolutionary-neural system, based on the SVM classifier. The proposed method shows good performance with high sensitivity (90.19%), specificity (99.39%), and accuracy (98.85%).