Documente Academic
Documente Profesional
Documente Cultură
he surface electrocardiogram (ECG) provides a noninvasive way to study atrial fibrillation (AF) mechanism(s) and to investigate the effects of remodeling
and treatment on AF. Time domain methods can be
used to characterize the signal in the surface ECG.
Analysis can be done through direct observation of the original signal or through methods used to obtain and analyze atrial
activity. These methods include cancellation techniques, vector analysis, autocorrelation, and combined time-frequency
domain methods to further investigate different atrial activity
parameters. Waveform characterization during AF can
improve our understanding of its mechanism(s) and can help
us learn about the potential clinical uses of the surface ECG.
AF is a supraventricular tachyarrhythmia that affects 0.4%
of the general population. The likelihood of developing AF
increases with age, less than 1% in those under 60 years of age
and greater than 6% in those over 80 years of age [1]. During
AF, the coordinated activation of the atria is replaced by an
uncoordinated atrial activation that leads to the deterioration of
atrial mechanical function with a consequent increased risk of
stroke and increased mortality.
There are a number of theories about the mechanism(s) of AF,
including rapid firing of foci from one or more pulmonary veins
or other sites [2], multiple reentrant wavelets circulating within
the atria [3], and wandering rotors [4]; however, the mechanisms of initiation, maintenance, and termination are not completely understood. Wijffels et al. showed that
electrophysiologic remodeling takes place in the atria during
AF, which supports the idea that AF is a progressive arrhythmia [5]. This phenomenon has been attributed to multiple factors including molecular, electrophysiologic, contractile, and
structural remodeling, each of which might contribute to
detectable features in the ECG. The surface ECG provides a
noninvasive way to study these mechanisms in humans. Some
of the advantages of using the surface ECG include the ability
to record data for a long period of time, the minimal cost and
risks involved compared with invasive electrophysiologic
study, and its reflection of the global activity in the atria and
ventricles during AF.
The surface ECG can be used for manual or automatic detection of AF [6]. It directly reflects the electrophysiological
processes that underlie AF, including refractory periods [5],
autonomic tone [7], drug effects [8], and linking [9]. Therefore,
it can be used to better understand the mechanisms of AF and
to study the effects of remodeling and the response to treatment. The surface ECG may also be used to predict the pattern
of occurrence of AF, the likelihood of termination or persistence, and the probability of recurrence in different patients.
One way to investigate the use of the surface ECG is by
employing time domain methods to characterize the signal.
The first step to characterize AF is through direct observations
from the original signal, so we will start this article by introducing the electrocardiographic characteristics of AF in the
surface ECG including ventricular rhythm observations during
AF. We will then describe some methods commonly used to
obtain and characterize atrial activity. These include obtaining
a remainder ECG in which only atrial activity is present,
investigating the spatial and temporal organization of AF, and
analyzing the atrial activity to learn about potential clinical
uses of the remainder ECG.
AF and Electrocardiographic Characteristics
One use of the surface ECG is for the diagnosis of AF. In the
surface ECG, AF is characterized by rapid atrial activity that is
irregular in timing and morphology. Discrete P-waves are absent
and replaced by an oscillating baseline that consists of lowamplitude fibrillatory waves (F-waves). The shape, amplitude,
and regularity of F-waves vary from patient to patient. In some
cases, the pattern of atrial activity can be similar to atrial flutter,
mostly regular and with high-amplitude F-waves, while in other
cases, it can be less regular, have lower amplitude, or both. Atrial
rates detected from the surface ECG in AF vary between 240
and 540 beats/min [10] with an average of 350 beats/min, and
there is a substantial overlap between atrial flutter and AF rates.
Distinguishing between AF and atrial flutter is clinically important since the treatment may be different for the two. Figure 1
illustrates some examples of AF and atrial flutter with regular
and irregular ventricular rhythm. The irregular ventricular
response may be a clue to the presence of AF, although this is a
consequence of this arrhythmia and is not necessarily present.
Ventricular Rhythm in AF
NOVEMBER/DECEMBER 2006
(a)
(b)
(c)
1 mV
(d)
1s
Fig. 1. AF and atrial flutter examples: (a) AF with irregular ventricular rhythm, (b) AF with paced (regular) ventricular rhythm,
(c) atrial flutter with regular ventricular rhythm, and (d) atrial flutter with irregular ventricular rhythm.
1.8
1.6
1.4
R-R Interval (s)
QRS detection schemes are available, which facilitate measurement of the R-R interval directly from the ECG in the time
domain without additional processing [11]. During AF, the RR interval and hence, the ventricular rate, is commonly
observed as irregular. However, ventricular activity of this sort
is present not only in AF but also in a variety of other arrhythmias, including multifocal atrial tachycardia and atrial flutter
with variable atrioventricular (AV) block. Conversely, AF
may be present with a regular ventricular rate, as in the case of
artificial ventricular pacing. Although irregular ventricular
activity is commonly associated with AF, it cannot function as
the sole diagnostic criterion. In fact, the presence of AF has
been shown to be under-recognized in paced patients, with
important adverse clinical consequences since the R-R intervals are regular [12].
In addition to problems associated with misinterpreting the
ECG based on ventricular activity, there is also a major hazard
associated with the ventricular response to AF in the presence
of one or more accessory pathways. These pathways facilitate
ventricular activity that is not mediated by the AV node, resulting in ventricular preexcitation. Therefore, in some patients, the
ventricular rate can become extremely rapid, with the risk of
ventricular fibrillation and sudden cardiac death [13].
The cause of the irregularity of the ventricular response during AF can be related to the irregularity of the atrial activity
itself as well as to irregularity imposed by the AV node.
Concealed conduction in the AV node takes place such that
many atrial impulses, though not conducted, influence the timing of the output to the ventricles. Bootsma et al. concluded
that during AF, the irregular sequence of variable strength atrial pulses approach the AV node from different directions,
resulting in an irregular ventricular response [14]. Because the
interval at which the AV node is activated is inversely related
to conduction time, an irregular ventricular response results.
Kirsh et al. have demonstrated that during AF, the AV node
transmits the irregular pattern of atrial excitation to the ventricles at a scaled-down rate, and no additional special properties
of the AV node need to be invoked to explain the irregularity of
the ventricular response [15]. It should be noted, however, that
autonomic nervous system input modifies both the electrophysi-
1.2
1
0.8
0.6
0.4
0.2
0
0
500
2,500
3,000
Fig. 2. R-R interval variability over 50 min. The first 1,400 R-R
intervals correspond to sinus rhythm, and the rest correspond
to AF. The R-R variability is much larger and the R-R intervals
are shorter for AF compared to sinus rhythm. These properties
of the R-R intervals help in the detection of AF.
NOVEMBER/DECEMBER 2006
25
cancellation technique to account for the variations in the electrical axis of the heart, variations that are primarily due to the
respiratory activity [22]. Other methods that take advantage of
multiple lead information have also been developed, such as
the two decomposition methods of principal component analysis and blind source separation [23], [24].
Langley et al. compared three of these techniques, namely,
spatiotemporal QRST cancellation, blind source separation,
and principal component analysis, and determined that there
were no significant differences between the atrial frequencies
estimated by each of these techniques [25].
Despite all the improvements in cancellation methods, there
was still concern that the residuals of ventricular activity and
other distortions due to cancellation may alter the characterization of F-waves. Xi et al. validated these methods by comparing gold-standard pure AF ECGs obtained by briefly stopping
pacing in patients with complete AV block after undergoing
AV junctional ablation with remainder ECGs obtained by cancellation of paced beats before and after the pure AF segments
[26]. Comparison between pure AF segments and QRST-canceled segments preceding and following the pure AF segments
showed similar characteristics in the frequency domain.
Leads II and V1 have the largest ratio of atrial to ventricular
signal amplitude and, therefore, are often chosen for analysis.
It is believed that subtraction has the best results in lead V1
resulting in high-amplitude F-waves, and this is due to the
leads proximity to the right atrium.
After obtaining a remainder ECG, F-wave characterization
is of interest. One parameter to be analyzed is F-wave amplitude. The distribution of amplitude over time can be also be
quantified by using the amplitude probability density function,
as is done for the intra-atrial leads [10]. Another goal is to
study the organization of the arrhythmia, and this can be done
by using vector analysis and methods of autocorrelation or
cross-correlation. Frequency analysis is often used in addition to
time domain analysis to determine atrial rate and to characterize
different patterns of occurrence of AF and the effects of drugs.
Cancellation Techniques
The analysis of atrial activity in the surface ECG is complicated by the simultaneous presence of ventricular activity, which
is typically of much greater amplitude. There are two ways to
investigate F-waves: isolate segments free of QRST complexes [17] where only F-waves are present or cancel the ventricular activity and obtain a remainder ECG that consists only of
atrial activity.
Slocum et al. introduced a template subtraction method that
takes advantage of the lack of a constant phase relationship
between atrial and ventricular activities and the mostly consistent amplitude and morphology of the QRST complexes [18].
In this method, ventricular beats are detected and aligned by
using the point of maximum upward or downward slope of the
QRS complex, also known as a fiducial point. A mean or
median beat is generated by aligning all detected ventricular
beats at their fiducial points, where the window length is
determined by the minimum or mean R-R interval. A template
of mean or median beats is constructed and subtracted from
the original signal, resulting in a remainder ECG. This process
is illustrated in Figure 3.
The remainder ECG consists mostly of dissociated atrial
activity, although QRST residuals and noise are often present.
Improvements to this cancellation method have been proposed. One method uses adaptive recurrent filtering [19] such
that the impulse response of the filter is adapted to successive
beats and then used for subtraction [20], [21]. In many cases,
the mean or median beat cannot represent each individual beat
accurately since QRST morphology is affected by respiration
and changes in the orientation of the heart caused by body
position changes. Stridh and Srnmo used a spatiotemporal
V1
TemplateV1
(a)
(b)
RemainderV1
(c)
Fig. 3. QRST cancellation: (a) Lead V1 during AF, (b) template of median beats created from lead V1, and (c) remainder ECG
obtained by subtracting the template from lead V1.
NOVEMBER/DECEMBER 2006
Vector Analysis
AF has been shown to also have quantifiable spatial organization [38][41]. Using intracardiac recordings from an orthogonal catheter, Gerstenfeld et al. [9] showed that AF activation
wavefronts in the right atrium have the tendency to follow the
paths of previous excitation, a phenomenon called transient
linking [41]. Botteron and Smith [40] described how spatial
organization could be quantified through cross-correlation of
multiple intracardiac recordings in relation to distance
between sites. Gray and others have observed and modeled AF
sustained by spiral waves and rotors [4].
Can spatial organization of AF be measured using the surface
ECG? The possibility exists if it can be assumed that
F-waves in the ECG accurately reflect actual atrial activation
directions and conduction velocities. To test this assumption, Ng
et al. initially studied ECG recordings of atrial flutter of known
mechanisms using vectorcardiography [42], [43]. Typical atrial
flutter reflects a macroreentrant circuit around the tricuspid annulus in the right atrium. When viewing the right atrium from the
left anterior oblique perspective, the activation sequence of the
circuit is more commonly counterclockwise around the tricuspid
annulus; however, clockwise activation is also possible. The vector loops of the flutter waves, as illustrated in Figure 4, proved to
be accurate reflections of the counterclockwise or clockwise
rotational direction of the atrial flutter circuits.
mV
0.05
0.1
Autocorrelation
Although AF is not a completely regular rhythm such as atrial flutter or atrial tachycardia, a quantifiable level of
temporal organization exists [36]. This
organization allows estimation of the
atrial rate using the autocorrelation
function. Slocum et al. used this
method to discriminate between AF and AV-dissociated Pwaves from the remainder ECG [37]. If the atrial activity is
periodic and AV dissociated, the autocorrelation of the
remainder ECG signal will contain peaks at the atrial cycle
length and multiples of the cycle length. The discrimination
between AF and AV dissociated P-waves was based on the
peak amplitude and duration of the autocorrelation function
in Leads II and V1. The best parameter for detecting AF
seemed to be peak amplitude, with a significantly higher
peak for AF recordings compared to the control group. The
best detection algorithm was achieved when amplitudes of
the peaks in leads II and V1 were added together and this
new parameter was used as a classifier. The properties of AF
that allow estimation of atrial rate using autocorrelation are
also represented by the narrow band characteristics of the
power spectrum in frequency domain.
0.05
0.1 0.05
0.05
0.05 0.1
(a)
0.05
mV
mV
(b)
Fig. 4. Surface ECG atrial flutter waves viewed from the left anterior oblique perspective with (a) counterclockwise and (b) clockwise reentrant circuits around the tricuspid annulus.
NOVEMBER/DECEMBER 2006
27
Cycle Number
3
4
5
II
V1
V2
0.1 mV
V3
V4
X
Y
X
Y
V5
V6
0
500
Surface ECG f Waves
(a)
ms
1,000
Fig. 5. Vector loops: (a) 1 s of pure AF signal divided into six segments of equal length and (b) the three-dimensional
vector loops of the individual estimated AF cycles.
NOVEMBER/DECEMBER 2006
The surface ECG is a convenient, cost effective, and noninvasive tool for the study of AF. It can be used to examine the
hypothesized mechanisms of AF as well as to quantify and
assess the effect of electrophysiological remodeling and the
effectiveness of treatment on different types of AF. Time
domain methods can be used to characterize the signal in the
surface ECG. We described observations that can be obtained
directly from the signal, such as the general characteristics of
AF in the surface ECG and the ventricular response to AF.
We also discussed commonly used methods to characterize
atrial activity. These methods include cancellation techniques,
vector analysis, and autocorrelation. We concluded with a
brief discussion about the success of combining time and frequency domain methods for a more thorough understanding
of the characteristics of the atrial activity in the surface ECG.
Whether the study of atrial activity in the surface ECG can be
used to distinctively distinguish between different mechanisms of AF is not yet known, but further investigation can
improve our understanding of these mechanisms and help
with the management of this common arrhythmia.
Acknowledgments
29
References
[1] V. Fuster, L.E. Ryden, R.W. Asinger, et al. ACC/AHA/ESC guidelines for the
management of patients with atrial fibrillation: Executive summary. A report of the
American College of Cardiology/American Heart Association task force on practice guidelines and the European Society of Cardiology Committee for practice
guidelines and policy conferences (committee to develop guidelines for the management of patients with atrial fibrillation) developed in collaboration with the
North American Society of Pacing and Electrophysiology, Circulation, vol. 104,
pp. 21182050, Oct. 2001.
[2] M. Haissaguerre, P. Jais, D.C. Shah, A. Takahashi, M. Hocini, G. Quiniou,
S. Garrigue, A. Le Mouroux, P. Le Metayer, and J. Clementy, Spontaneous initiation of atrial fibrillation by ectopic beats originating in the pulmonary veins, New
Eng. J. Med., vol. 339, pp. 659666, Sept. 1998.
[3] G.K. Moe, On the multiple wavelet hypothesis of atrial fibrillation, Arch. Int.
Pharmacodyn. Ther., vol. 140, pp. 183188, 1962.
[4] J. Jalife, O. Berenfeld, and M. Mansour, Mother rotors and fibrillatory
conduction: A mechanism of atrial fibrillation, Cardiovasc. Res., vol. 54,
pp. 204216, May 2002.
[5] M.C. Wijffels, C.J. Kirchhof, R. Dorland, and M.A. Allessie, Atrial fibrillation begets atrial fibrillation. A study in awake chronically instrumented goats,
Circulation, vol. 92, pp. 19541968, Oct. 1995.
[6] J. Slocum, A.V. Sahakian, and S. Swiryn, Diagnosis of atrial fibrillation from
surface electrocardiograms based on computer-detected atrial activity, J.
Electrocardiol., vol. 25, pp. 18, Jan. 1992.
[7] M. Zimmermann, Autonomic tone and atrial fibrillation, J. Cardiovasc.
Electrophysiol., vol. 14, pp. 559564, June 2003.
[8] A. Bollmann, N.K. Kanuru, K.K. McTeague, P.F. Walter, D.B. DeLurgio,
and J.J. Langberg, Frequency analysis of human atrial fibrillation using the surface electrocardiogram and its response to ibutilide, Amer. J. Cardiol., vol. 81,
pp. 14391445, June 1998.
[9] E.P. Gerstenfeld, A.V. Sahakian, S. Swiryn, Evidence for transient linking of
atrial excitation during atrial fibrillation in humans, Circulation, vol. 86,
pp. 375382, Aug. 1992.
[10] J. Slocum, A.V. Sahakian, and S. Swiryn, Computer discrimination of atrial
fibrillation and regular atrial rhythms from intra-atrial electrograms, Pacing Clin.
Electrophysiol., vol. 11, pp. 610621, May 1988.
[11] B.U. Khler, C. Hennig, and R. Orglmeister, The principles of software QRS
detection: Reviewing and comparing algorithms for detecting this important ECG
waveform, IEEE Eng. Med. Biol. Mag., vol. 21, pp. 4257, Jan./Feb. 2002.
[12] A.M. Patel, D.C. Westveer, K.C. Man, J.R. Stewart, and H.I. Frumin,
Treatment of underlying atrial fibrillation: Paced rhythm obscures recognition, J.
Amer. Coll. Cardiol., vol. 36, pp. 784787, Sept. 2000.
[13] E.N. Prystowsky, D.W. Benson, Jr., V. Fuster, R.G. Hart, G.N. Kay,
R.J. Myerburg, G.V. Naccarelli, and D.G. Wyse, Management of patients with
atrial fibrillation. A statement for healthcare professionals. From the subcommittee
on electrocardiography and electrophysiology, American Heart Association,
Circulation, vol. 93, pp. 12621277, Mar. 1996.
[14] B.K. Bootsma, A.J. Hoelsen, J. Strackee, and F.L. Meijler, Analysis of R-R
intervals in patients with atrial fibrillation at rest and during exercise, Circulation,
vol. 41, pp. 783794, May 1970.
[15] J.A. Kirsh, A.V. Sahakian, J.M. Baerman, and S. Swiryn, Ventricular
response to atrial fibrillation: Role of the atrioventricular conduction pathways, J.
Amer. Coll. Cardiol., vol. 12, pp. 12651272, Nov. 1988.
[16] M.M. Gallagher, K. Hnatkova, F.D. Murgatroyd, J.E. Waktare, X. Guo, A.J.
Camm, and M. Malik, Evolution of changes in the ventricular rhythm during
paroxysmal atrial fibrillation, Pacing Clin. Electrophysiol., vol. 21,
pp. 24502454, Nov. 1998.
[17] D.S. Rosenbaum and R.J. Cohen, Frequency based measures of atrial fibrillation in man, in Proc. Annu. Int. Conf. IEEE Engineering Medicine Biology
Society, 1990, vol. 12, pp. 582583.
[18] J. Slocum, E. Byrom, L. McCarthy, A.V. Sahakian, and S. Swiryn,
Computer-detection of atrioventricular dissociation from surface electrocardiograms during wide QRS complex tachycardias, Circulation, vol. 72,
pp. 10281036, Nov. 1985.
[19] A.V. Sahakian and K.H. Kuo, Canceling the cardiogenic artifact in impedance pneumography, in Proc. Annu. Int. Conf. IEEE Engineering Medicine
Biology Society, 1985, vol. 7, pp. 855859.
[20] N.V. Thakor and Y.S. Zhu, Applications of adaptive filtering to ECG analysis: Noise cancellation and arrhythmia detection, IEEE Trans. Biomed. Eng.,
vol. 38, pp. 785794, Aug. 1991.
[21] P. Laguna, R. Jane, E. Masgrau, and P. Caminal, The adaptive linear combiner with a periodic-impulse reference input as a linear comb filter, Signal
Process., vol. 48, pp. 193203, 1996.
[22] M. Stridh and L. Srnmo, Spatiotemporal QRST cancellation techniques for
analysis of atrial fibrillation, IEEE Trans. Biomed. Eng., vol. 48, pp. 105111,
Jan. 2001.
[23] P. Langley, J.P. Bourke, and A. Murray, Frequency analysis of atrial fibrillation, Comput. Cardiol., vol. 27, pp. 6568, 2000.
[24] J.J. Rieta, F. Castells, C. Sanchez, V. Zarzoso, and J. Millet, Atrial activity
extraction for atrial fibrillation analysis using blind source separation, IEEE
Trans. Biomed. Eng., vol. 51, pp. 11761186, July 2004.
[25] P. Langley, M. Stridh. J.J. Rieta, L. Srnmo, J. Millet-Roig, A. Murray,
Comparison of atrial rhythm extraction techniques for the estimation of the main
NOVEMBER/DECEMBER 2006