Investigating Feature Selection and Random Forests for Inter-Patient Heartbeat Classification

Jose Francisco Saenz-Cogollo,Maurizio Agelli

doi:10.3390/a13040075

Abstract

Finding an optimal combination of features and classifier is still an open problem in the development of automatic heartbeat classification systems, especially when applications that involve resource-constrained devices are considered. In this paper, a novel study of the selection of informative features and the use of a random forest classifier while following the recommendations of the Association for the Advancement of Medical Instrumentation (AAMI) and an inter-patient division of datasets is presented. Features were selected using a filter method based on the mutual information ranking criterion on the training set. Results showed that normalized beat-to-beat (R–R) intervals and features relative to the width of the ventricular depolarization waves (QRS complex) are the most discriminative among those considered. The best results achieved on the MIT-BIH Arrhythmia Database were an overall accuracy of 96.14% and F1-scores of 97.97%, 73.06%, and 90.85% in the classification of normal beats, supraventricular ectopic beats, and ventricular ectopic beats, respectively. In comparison with other state-of-the-art approaches tested under similar constraints, this work represents one of the highest performances reported to date while relying on a very small feature vector.

Highlights

The electrocardiogram (ECG) is the register of electrical potentials, variable over time, produced by the myocardium during the cardiac cycle
According to the standard ANSI/Advancement of Medical Instrumentation (AAMI) EC57:1998/(R)2008, the methods that perform heartbeat classification should at most distinguish between the following classes: normal beats (NB), supraventricular ectopic beats (SVEB), ventricular ectopic beats (VEB), fusion beats (FB), and unclassifiable or paced beats
Time-domain features extracted from the single-lead ECG were objectively selected by their information content, and the heartbeat classification performance using Random Forests (RF) was fairly evaluated by following the AAMI guidelines and the inter-patient paradigm

Summary

Introduction

The electrocardiogram (ECG) is the register of electrical potentials, variable over time, produced by the myocardium during the cardiac cycle. Thanks to the availability of pocket or wearable devices for single lead ECG recording, it has become possible to perform the acquisition of the ECG signal even in ambulatory contexts for applications of prevention and risk management. We focus our attention on the problem of classifying single heartbeats and separating the normal electrical activity originated in the sinus node from the ectopic activity originated elsewhere. According to the standard ANSI/AAMI EC57:1998/(R)2008, the methods that perform heartbeat classification should at most distinguish between the following classes: normal beats (NB), supraventricular ectopic beats (SVEB), ventricular ectopic beats (VEB), fusion beats (FB), and unclassifiable or paced beats The successful application and dissemination of such approaches require the development of reliable, and lightweight algorithms for the automatic detection and classification of signal anomalies in order to reduce the amount of data and the number of events needed to be sent to the physician for making a proper risk assessment and giving advice [1].

Methods

Results

Discussion

Conclusion