ECG ARTEFACT DETECTION ALGORITHM

An Algorithm to Improve Long-term ECG Analysis

Susana Brás

1,2

, Nuno Ferreira

and João Paulo Silva Cunha

1,2

Institute of Electronics and Telematics Engineering of Aveiro (IEETA), Aveiro University

Campus Universitário de Santiago, Aveiro, Portugal

Departamento de Engenharia Electrónica, Telecomunicações e Informática, Universidade de Aveiro

Campus Universitário de Santiago, Aveiro, Portugal

Biodevices S.A., Rua 5 de Outubro, Porto, Portugal

Keywords: Electrocardiogram, Artefact detection, Threshold analysis.

Abstract: Newly devices allow the analysis and collection of very long-term electrocardiogram (ECG). However,

associated with this devices and long-term signal, are artefacts that conduce to misleading interpretations

and diagnosis. So, new developments over automatic ECG classification are needed for a reliable

interpretation. The feasibility of the cardiac systems is one of the main concerns, once they are currently

used as diagnosis or help systems. In this project, an artefact detection algorithm is developed, dividing the

time-series in intervals of signal and artefact. The algorithm is based on the assumption that, if the analysed

frame is signal, there is not an abrupt alteration over consecutive short windows. So, the time-series is

divided in consecutive nonoverlapped short windows. Over these windows, it is calculated the time-series

standard deviation, the maximum and minimum slope. A threshold-based rule is applied, and the algorithm

reveals that, in mean, it is verified a 99.29% of correctly classified signal and only 0.71% of signal

erroneously classified. Over the results obtained, the algorithm seems to present good results, however it is

needed its validation in a wider and representative sample with segments marked as artefact by multiple

specialists.

1 INTRODUCTION

The electrocardiogram (ECG) studies started with

visual inspection of the wave morphology alterations

(Malik, 2003). Due to the introduction of long-term

ECG devices, the visual inspection became time

consuming. To address this shortcoming, automatic

systems and algorithms were presented for a fast and

efficient analysis. The scientific community

proposed algorithms for the delineation and

identification of the complexes that compose the

heartbeat (P, QRS and T) (Pan, 1985); (Hamilton,

1986); (Almeida, 2010); (Vila, 2000); (Martinez,

2004). The analysis of different ECG characteristics

allowed the evaluation of specific pathologies, e.g.,

arrhythmias (Chin, 2010); (Tsipouras, 2002); or

ventricular repolarization abnormalities (Malik,

2003).

All the advances on the ECG analysis and

classification introduce a high necessity of new

developments in this area to obtain the most feasible

results. In rest ECG analysis (Stern, 1975), the

patient cardiac function is evaluated, studying the

alterations to normal under a resting condition. In

this kind of analysis, the ECG time-series is almost

noise free. In Holter analysis (Gibson, 2007),

physicians are interested in the analysis of long-term

ECG. The purpose is the evaluation of cardiac

function during daily routine activities, usually in a

24h or 48h exam. Since the Holter is a long-term

exam with the goal to evaluate specific alterations or

abnormalities, the presence of artefacts could mask

important events. Nowadays, very long-term ECG

monitoring is used as telemetric medicine (Mittal,

2011), or for online and real-time evaluation of

patient cardiac function. The Vital Responder

project is an example of such long-term ECG

monitoring. The main goal of the project is to

develop a system able to identify fatigue/stress,

during firefighters daily routine activities, focusing

on the cardiovascular analysis. Between first

responder professionals, firefighters registered the

329

Brás S., Ferreira N. and Silva Cunha J..

ECG ARTEFACT DETECTION ALGORITHM - An Algorithm to Improve Long-term ECG Analysis.

DOI: 10.5220/0003729503290333

In Proceedings of the International Conference on Bio-inspired Systems and Signal Processing (BIOSIGNALS-2012), pages 329-333

ISBN: 978-989-8425-89-8

 2012 SCITEPRESS (Science and Technology Publications, Lda.)

highest number of deaths on duty; some of them

associated with cardiac complications, which could

be consequence of exposure to stressful events.

The ECG based algorithms sometimes produce

misleading interpretation due to artefacts presented

in the time-series. Thus, algorithm for artefact

detection is of upmost importance that will allow

rejecting anomalous information, which induces in

error the automatic algorithms. The goal of this

paper is to present new algorithm for artefact

detection for very long-term ECG monitoring.

Basically, the algorithm evaluates the alterations of

the wave over consecutive windows. The entire

algorithm development was focused on the

assumption that ECG time-series have not abrupt

alterations between consecutive windows, if they are

considered sufficiently shorter. The long term ECG

could present differences along the exam, however,

they are gradually inspected. Therefore, if the

windows are sufficiently short, i.e., 2 or 3 seconds,

the alterations should not be significant if no artefact

is present.

In section 2, the Methods are described divided

in: Evaluated Database characteristics, ECG Pre-

processing, Artefact Detection Algorithm and the

Performance Evaluation. The Results are presented

in section 3 and in section 4 conclusions are drawn.

2 METHODS

The ECG was collected using the VitalJacket®

(Figure 1). The VitalJacket® is a wearable very

long-term ECG monitoring device composed by

microelectronics embedded into the textile

manufactured in the form of a simple t-shirt (Cunha,

2010). The VitalJacket® is a confortable t-shirt and

consequently does not collide with mobility. It can

record 5 lead ECG and accelerometry during up to 5

days in a single battery charge. It is equipped with a

memory card, where the data is recorded. Also, the

communication by Bluetooth in real time to a

computer or smartphone is possible, allowing the

real time analysis of data and algorithms

implementation. Biodevices S.A. has successfully

concluded the certification process according to the

standards ISO9001 and ISO13485 and the approval

of Vital Jacket® as a medical “Ambulatory ECG

device” according to the MDD directive 42/93/CE

that regulates medical devices in Europe. Vital

Jacket® has been granted with the CE1011 mark

(Cunha, 2010).

Figure 1: Vital Jacket, the wearable very long-term ECG

recording device used in the study.

2.1 Evaluated Database

The data was collected from 8 individual divided in

segments of noise and signal. The segments of

consecutive epochs were validated by a senior

cardiologist. Over the very long-term ECG data, the

specialist chose ECG segments where the time series

is not corrupted by artefact. On the other hand, there

are also chosen segments of the time series, clearly

marked by artefacts. Since, sometimes, the artefact

is not persistent, in some of the cases was not

possible to obtain a segment only marked by

artefact. So, it is important to point out that the data

contained in the signal segment is free of artefacts;

nevertheless the data in the noise segment is not

100% artefact.

The algorithm was validated not in all collected

signal, but in the segments chosen by the specialist,

and described in table 1. The signal (noise) segments

correspond to 0.25% (0.10%) of the total exams.

Since the data was collected in a non-controlled

scenario, it is contaminated by artefacts. Therefore,

consecutive segments of signal (noise) in each exam

have small duration.

Table 1: Database characteristics, the data was collected

using the VitalJacket®. Fs is the sampling frequency, h

corresponds to hours, m to minutes and s to seconds. The

segments are constituted by consecutive epochs of the

ECG long-term exam.

Example

(Hz)

Exam

duration

(h:m:s)

Signal

segment

(h:m:s)

Noise

segment

(m:s)

1 500