Detection of heart diseases using PCG signals

Abstract

In Phonocardiography (PCG), the heart sound signals are recorded using an electronic stethoscope and are displayed graphically on the PC/laptop rather than listening the heart sounds as done in traditional auscultation. More details are accessible visually, because the analysis is not limited by the human audibility range or experience of the physician while listening. The heart diseases can be detected even before the symptoms of pathology appear and this makes it a highly potential diagnostic test for the future. The aim of this study is to detect various heart diseases using the PCG signals. Many diagnostic features can be extracted using PCG which otherwise require tests like Electrocardiography (ECG) or Echocardiography. This can save a lot of money and time. Moreover, the requirement of instruments is minimum i.e. an electronic stethoscope and a PC/laptop which are available even at primary health care centres. This will increase the outreach of diagnosis to under-privileged and rural people. In order to detect murmurs efficiently, two new features have been proposed. A total of 23 features are evaluated for normal and murmur signals. Then 5 optimal features are finally selected for classification using Ranker and Info Gain Evaluation method for feature reduction. Finally, an algorithm using Naïve Bayes classifier is proposed for murmur detection and arrhythmia detection (like Tachycardia or Bradycardia).