ATRIAL FIBRILLATION

Atrial Fibrillation and

Waveform Characterization
A Time Domain Perspective in the Surface ECG
BRAND X PICTURES

BY SIMONA PETRUTIU, JASON NG,

GRACE M. NIJM, HAITHAM AL-ANGARI,
STEVEN SWIRYN, AND
ALAN V. SAHAKIAN

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],

24 IEEE ENGINEERING IN MEDICINE AND BIOLOGY MAGAZINE

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

The R-R interval in the surface ECG represents the time

between ventricular electrical systoles. Currently, numerous
0739-5175/06/$20.002006IEEE

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.

IEEE ENGINEERING IN MEDICINE AND BIOLOGY MAGAZINE

ological function of the AV node and the atria themselves.

Decreased vagal tone or increased sympathetic stimulation
results in increased conduction through the AV node [13].
The characteristics of the R-R interval vary from the onset
of AF to its termination. The variation present during AF is
much greater than the variation typically present in sinus
rhythm. Figure 2 shows an example of R-R intervals over
25 min of sinus rhythm and 25 min of AF, and the greater
variability during AF can be observed. Since paroxysmal AF
involves episodes of AF that self-terminate, it is reasonable to
analyze the R-R intervals of these entire episodes. Gallagher et
al. observed that the mean R-R interval and the variability of
the R-R intervals increase from the onset of AF [16]. In addition, they found that immediately before termination, the duration and variability of the R-R interval increase. This increase
in the interval may indicate increased parasympathetic

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

1,000 1,500 2,000

Interval Number

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

influence at the AV node. Another possible interpretation of

the change in the pattern of the ventricular response before termination may be related to the changes occurring in the atria
prior to termination.
Although the study of ventricular rate can be a clue in the
detection of AF, in order to better understand the mechanisms
of AF and the effects of drugs or autonomic tone on the
arrhythmia, it is necessary to study the characteristics of the
atrial fibrillatory waves themselves.

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

Fibrillatory Wave Amplitude

The amplitude of F-waves is often described as fine or coarse,

traditionally defined as less than or greater than 50 V,

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.

26 IEEE ENGINEERING IN MEDICINE AND BIOLOGY MAGAZINE

NOVEMBER/DECEMBER 2006

respectively. The amplitude of F-waves in AF is lower when

compared to the amplitude of P-waves during sinus rhythm
[27]. However, within the same patient, there is a correlation
between the amplitude of the signal during sinus rhythm and
during AF at the same recording site [28].
Xi et al. described one way to quantify F-wave amplitude
[29]. First, the region of QRS complexes was excluded from
analysis to avoid contribution of QRS residuals. The four
individual F-waves with the largest peak-to-peak amplitude
were selected, and the average of these four amplitudes was
selected as the representative peak-to-peak amplitude.
Another way to quantify amplitude is by looking at the average value of the peak-to-peak F-wave amplitude for the
entire duration of the recording.
F-wave amplitude has been linked to specific heart conditions like rheumatic heart disease and arteriosclerotic heart
disease [30]. Culler et al. introduced the criteria of very
coarse AF as greater than 0.25 mV versus straight-line,
barely visible F-waves [31]. Both studies associated
rheumatic heart disease with coarse F-waves and arteriosclerotic heart disease with fine F-waves. However,
Morganroth et al. found no difference in F-wave amplitude
between rheumatic and nonrheumatic AF and no significant
correlation between F-wave amplitude and echocardiographic left atrial size [32]. Similar contradicting conclusions were obtained in studies relating F-wave amplitude to
left atrial size and left atrial appendage function [33], [34].
In order to characterize atrial activity using fibrillatory
wave amplitude and relate it to the mechanisms and pathophysiology of AF, there is an implicit assumption of stability over time. Xi et al. showed that F-wave amplitude has
high intrapatient repeatability from ECG to ECG over a
period of 24 hours, whereas the interpatient variability is
high [29]. Similar results were found for atrial rate. The
relationships of AF pattern, antifibrillatory drugs, and age to
F-wave amplitude were also investigated [35]. There was no
statistical significance between F-wave amplitude and different patterns of AF, which include paroxysmal, persistent,
and permanent. F-wave amplitude was also not different for
patients taking antifibrillatory drugs compared with those
not taking antifibrillatory drugs, and there was also no association of age with F-wave amplitude.
There are certain factors that make it
difficult to relate F-wave amplitude to
specific etiologies of AF or atrial size.
mV
Some of these include the differences in
the thickness of the constituent tissues
0.1
of the chest wall between different
patients and the skin-electrode inter0.05
face. This could explain why many
0
studies offered conflicting results about
the relationship of F-wave amplitude
0.05
and different etiologies of AF.

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.

IEEE ENGINEERING IN MEDICINE AND BIOLOGY MAGAZINE

NOVEMBER/DECEMBER 2006

27

Similar vectorcardiography methods were then applied to

AF [44]. AF signals were obtained during brief periods of
asystole when pacing was stopped (ranging from 2.14.8 s).
The AF cycle length of each signal was estimated using autocorrelation. The signal was then divided into segments with
lengths equal to the estimated cycle length. The three-dimensional vector loops were characterized by the planes that best
contained the loops and then categorized based on the direction of the planes (illustrated in Figure 5). An AF signal was
considered spatially organized if it had 30% or more of its
vector loops with planes within 30 of each other (a less than
0.03 probability to occur by chance). Fifteen of the 22 examples studied showed high organization with this criterion. Of
these 15, 12 were organized near the sagittal plane.
The sequences of the vector loop planes were also analyzed
to see if transient linking could be observed from the surface
ECG [45]. It was found that ten of the 22 AF signals had
episodes of five or more consecutive cycles that had planes
within 30 of each other. This suggests that transient linking is
a global and not just a local phenomenon as previously found.
The linking episodes were also more commonly near the sagittal plane (eight of ten).
There are a few possible explanations for the observed consistency of vector loop planes. One is that anatomic boundaries in
the right atrium such as the crista terminalis and the tricuspid
annulus tend to funnel activation wavefronts in the same direction
[46]. Another is that the vector loops reflect posterior-to-anterior
activation initiated by a focus from the posterior left atrium. It has
been shown that pulmonary vein triggers are often responsible for
the genesis and maintenance of AF [2]. A third explanation is that
the vector loops reflect a stable mother rotor that sustains AF.
Further study is needed to determine if surface ECG vector analysis could be used to differentiate AF mechanisms.
Combined Time and Frequency Domain Methods

In analyzing the atrial activity during AF, it has become

increasingly apparent that the most information is obtained by
using a combination of time and frequency domain methods.
The steps of generating the remainder ECG are performed in
the time domain, but additional information can be obtained
by further investigation of the remainder F-waves in the frequency domain.

Cycle Number
3
4
5

This combination of methods has proven very useful in the

diagnosis of AF from the surface ECG [6]. After generating a
remainder ECG, a power spectrum can be calculated.
Information in the power spectrum can be used to discriminate
between AF and other nonfibrillatory rhythms. For AF, most
of the power is concentrated in the 49 Hz band of the power
spectrum. Holm et al. showed that peak frequency (frequency
of maximum power) is a robust measure of intra-atrial cycle
length with the closest correlation found between Lead V1 and
right atrial electrograms [47].
The surface ECG represents a globally recorded signal. In
the case of AF, since the atria are not activated simultaneously,
the surface ECG may be different from locally recorded signals. Slocum and Ropella demonstrated that there is a correlation between the median frequency in the surface ECG Leads
II and V1 and simultaneous intracardiac recordings [48].
Other researchers have used similar time and frequency
domain methods to study the effects of antiarrhythmic drugs
on AF [49]. It has also been shown that F-waves correlate with
AF duration and AF recurrence [50].
Limitations of AF Analysis from the Surface ECG

During AF, the atria are being activated in an uncoordinated and

noncoherent manner. The signals recorded in the surface ECG
during AF are globally recorded signals and are therefore very
different from locally recorded signals. This makes it difficult to
infer much about local activations and spatial differences during
AF. It is not well understood how activity from different parts
of the atria contributes to the surface ECG. Most time domain
methods use lead V1 for analysis, which is dominated by right
atrial activity due to its proximal location to the right atrium.
Therefore it may be difficult to study left atrial events during
AF. Conversely, intracardiac signals typically obtained from
catheters have the opposite problem of having a myopic field of
view, unless a large number of recording sites are used.
Therefore, the surface ECG and intracardiac electrograms may
be best used simultaneously in the study of AF mechanisms, as
they provide complementary information to each other.
The low amplitude of the atrial activity in the surface ECG
presents a challenge in analyzing F-wave characteristics, especially in the presence of artifact and noise. Since the surface
ECG signal also depends on interpatient differences in body

II
V1
V2

0.1 mV

V3

V4

X
Y

X
Y

V5
V6
0

500
Surface ECG f Waves
(a)

ms
1,000

Derived f Wave Vector Loops

(b)

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.

28 IEEE ENGINEERING IN MEDICINE AND BIOLOGY MAGAZINE

NOVEMBER/DECEMBER 2006

characteristics and is affected by other physiologic parameters,

it is difficult to understand the influences of pathophysiology
on F-wave amplitude. This is reflected in the existence of
many conflicting results when trying to relate F-wave amplitude with different etiologies of AF.
Conclusions

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

This work was supported in part by a grant from The Dr.

Scholl Foundation.
Simona Petrutiu earned her B.S. and M.S.
degrees in computer engineering from
Northwestern University, Evanston,
Illinois, in 2002 and 2004, respectively.
She is currently pursuing her Ph.D. degree
in electrical engineering at Northwestern
University. Her research interests include
biomedical signal processing, cardiac
arrhythmias, and medical instrumentation.
Jason Ng is currently a research assistant
professor at the Feinberg School of Medicine
at Northwestern University, Evanston,
Illinois. He received his B.S. in electrical
engineering and computer engineering in
1998 and his M.S. and Ph.D. in electrical
engineering in 2000 and 2004, respectively,
all at Northwestern University. His research
interests include electrical mapping of atrial
arrhythmias and signal analysis of the surface electrocardiogram.
Grace M. Nijm earned her B.S. in computer engineering from Illinois Institute of
Technology and her B.S. in computer science from Benedictine University, both in
2004. In 2005, she was awarded the
National Science Foundation Graduate
Research Fellowship. She is currently pursuing her Ph.D. in electrical engineering at
Northwestern University.
IEEE ENGINEERING IN MEDICINE AND BIOLOGY MAGAZINE

Haitham M. Al-Angari received the B.S.

degree in electrical engineering in 1996
from King AbdulAziz University, Jeddah,
Saudi Arabia. In 2001, he received the
M.S. degree in biomedical engineering
from Northwestern University, Evanston,
Illinois. He is currently working toward
the Ph.D. degree in electrical engineering
at Northwestern University. His research interests are in
biomedical digital signal processing, with a special focus on
sleep apnea.
Steven Swiryn received the B.A. degree in
psychology from the University of
Michigan, Ann Arbor, in 1968 and the
M.D. degree from the University of Illinois,
Chicago, in 1973. He continued his graduate medical training at the University of
Illinois Hospitals and trained in clinical cardiac electrophysiology under the late Dr.
Kenneth M. Rosen. He is currently a professor of medicine,
Northwestern University Medical School, and adjunct professor of electrical engineering and computer science, Robert R.
McCormick School of Engineering and Applied Science,
Northwestern University, and he practices clinical cardiology
with HeartCare Midwest in Peoria, Illinois. He is board certified in internal medicine, cardiovascular disease, and in clinical cardiac electrophysiology. His research interests include
cardiac arrhythmia and conduction disease, especially AF
mechanisms, treatment, and detection and techniques of multichannel cardiac mapping.
Alan V. Sahakian received the Ph.D.
degree in electrical engineering from the
University of Wisconsin, Madison, in
1984. Since then he has been on the faculty of Northwestern University, Evanston,
Illinois, where he is currently professor of
biomedical engineering and professor and
program chair of electrical engineering
and computer science. He is also a member of the academic
affiliate staff of Evanston Hospital. He was a senior electrical engineer at Medtronic, Inc., and has served as a resident
visiting scholar at the Air Force Institute of Technology. His
research is in the general area of cardiac arrhythmia mechanisms and automatic diagnosis and therapy, with a special
focus on the electrophysiology and automatic treatment of
the atrial cardiac arrhythmias. He is also interested in smallscale magnetic resonance imaging and microwave/millimeter-wave measurement and imaging systems. His research is
funded by the National Science Foundation, the Dr. Scholl
Foundation, and Medtronic, Inc. He has coauthored more
than 150 papers, abstracts, patents, and book chapters. He
has served on the IEEE Engineering in Medicine and
Biology Society AdCom and as vice president for
Publications and Technical Activities.
Address for Correspondence: Alan V. Sahakian,
Department of Electrical Engineering and Computer Science,
Northwestern University, L356 Technological Institute, 2145
Sheridan Road, Evanston, IL 60208-3118 USA. E-mail:
sahakian@ece.northwestern.edu.
NOVEMBER/DECEMBER 2006

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

30 IEEE ENGINEERING IN MEDICINE AND BIOLOGY MAGAZINE

atrial frequency from the 12-lead electrocardiogram in atrial fibrillation, Comput.

Cardiol., vol. 29, pp. 2932, 2002.
[26] Q. Xi, A.V. Sahakian, and S. Swiryn, The effect of QRS cancellation on atrial
fibrillatory wave signal characteristics in the surface electrocardiogram, J.
Electrocardiol., vol. 36, pp. 243249, July 2003.
[27] J. Jenkins, K.H. Noh, A. Guezennec, T. Bump, and R. Arzbaecher, Diagnosis
of atrial fibrillation using electrograms from chronic leads: Evaluation of computer
algorithms, Pacing Clin. Electrophysiol., vol. 11, pp. 622631, May 1988.
[28] M.A. Wood, P. Moskovljevic, B.S. Stambler, and K.A. Ellenbogen,
Comparison of bipolar atrial electrogram amplitude in sinus rhythm, atrial fibrillation, and atrial flutter, Pacing Clin. Electrophysiol., vol. 19, pp. 150156, Feb. 1996.
[29] Q. Xi, A.V. Sahakian, J. Ng, and S. Swiryn, Atrial fibrillatory wave characteristics on surface electrogram: ECG to ECG repeatability over twenty-four hours
in clinically stable patients, J. Cardiovasc. Electrophysiol., vol. 15, pp. 911917,
Aug. 2004.
[30] M. Thurmann and J.G. Janney Jr., The diagnostic importance of fibrillatory
wave size, Circulation, vol. 25, pp. 991994, June 1962.
[31] M.R. Culler, J.A. Boone, and P.C. Gazes, Fibrillatory wave size as a clue to
etiological diagnosis, Amer. Heart J., vol. 66, pp. 435436, Sept. 1963.
[32] J. Morganroth, L.N. Horowitz, M.E. Josephson, and J.A. Kastor,
Relationship of atrial fibrillatory wave amplitude to left atrial size and etiology of
heart disease. An old generalization re-examined, Amer. Heart J., vol. 97,
pp. 184186, Feb. 1979.
[33] M.H. Aysha and A.S. Hassan, Diagnostic importance of fibrillatory wave
amplitude: A clue to echocardiographic left atrial size and etiology of atrial fibrillation, J. Electrocardiol., vol. 21, pp. 247251, Aug. 1988.
[34] A. Bollmann, K.H. Binias, K. Sonne, F. Grothues, H.D. Esperer, P. Nikutta,
and H.U. Klein, Electrocardiographic characteristics in patients with nonrheumatic
atrial fibrillation and their relation to echocardiographic parameters, Pacing Clin.
Electrophysiol., vol. 24, pp. 15071513, Oct. 2001.
[35] Q. Xi, A.V. Sahakian, T.G. Frohlich, J. Ng, and S. Swiryn, Relationship
between pattern of occurrence of atrial fibrillation and surface electrocardiographic
fibrillatory wave characteristics, Heart Rhythm, vol. 1, pp. 656663, Dec. 2004.
[36] A.C. Skanes, R. Mandapati, O. Berenfeld, J.M. Davidenko, J. Jalife,
Spatiotemporal periodicity during atrial fibrillation in the isolated sheep heart,
Circulation, vol. 98, pp. 12361248, Sept. 1998.
[37] J. Slocum, Use of the autocorrelation function to detect atrial fibrillatory
activity on the surface electrocardiogram, in Proc. Annu. Int. Conf. IEEE
Engineering Medicine Biology Society, 1991, vol. 13, no. 2, pp.732733.
[38] A.V. Sahakian, K.M. Ropella, J.M. Baerman, and S. Swiryn, Measuring the
organization of cardiac rhythms using the magnitude-squared coherence function,
IEEE Eng. Med. Biol. Mag., vol. 9, no. 1, pp. 2528, 1990.
[39] H.J. Sih, A.V. Sahakian, C.E. Arentzen, and S. Swiryn, A frequency domain
analysis of spatial organization of epicardial maps, IEEE Trans. Biomed. Eng.,
vol. 42, pp. 718727, July 1995.
[40] G.W. Botteron and J.M. Smith, Quantitative assessment of the spatial organization of atrial fibrillation in the intact human heart, Circulation, vol. 93,
pp. 513518, Feb. 1996.
[41] M.H. Lehmann, S. Denker, R. Mahmud, A. Addas, and M. Akhta, Linking:
A dynamic electrophysiologic phenomenon in macroreentry circuits, Circulation,
vol. 71, pp. 254265, Feb. 1985.
[42] J. Ng, A.V. Sahakian, W.G. Fisher, and S. Swiryn, Discrimination of atrial
flutter mechanisms from vector analysis of surface electrocardiograms, presented
at North Amer. Soc. Pacing Electrophysiology, Washington, DC, 2003.
[43] J. Ng, A.V. Sahakian, W.G. Fisher, and S. Swiryn, Atrial flutter loops
derived from the surface ECG: Does the plane of the loop correspond anatomically
to the macro-reentrant circuit?, J. Electrocardiol., vol. 36, pp. 181186, 2003.
[44] J. Ng, A.V. Sahakian, W.G. Fisher, and S. Swiryn, Further evidence for transient linking of atrial activity from vector analysis of surface electrocardiograms
recorded during atrial fibrillation, presented at the Heart Rhythm Soc., San
Francisco, CA, 2004.
[45] J. Ng, A.V. Sahakian, W.G. Fisher, and S. Swiryn, Surface ECG vector characteristics of organized and disorganized atrial activity during atrial fibrillation, J.
Electrocardiol., vol. 37, pp. 9197, 2004.
[46] F.X. Roithinger, A SippenGroenewegen, M.R. Karch, P.R. Steiner, W.S. Ellis,
and M.D. Lesh, Organized activation during atrial fibrillation in man: Endocardial
and electrocardiographic manifestations, J. Cardiovasc. Electrophysiol., vol. 9,
pp. 451461, May 1998.
[47] M. Holm, S. Pehrson, M. Ingemansson, L. Srnmo, R. Johansson, L. Sandhall,
M. Sunemark, B. Smideberg, C. Olsson, and S.B. Olsson, Non-invasive assessment of the atrial cycle length during atrial fibrillation in man: Introducing, validating and illustrating a new ECG method, Cardiovasc. Res., vol. 38,
pp. 6981, Apr. 1998.
[48] J. Slocum and K. Ropella, Correspondence between the frequency domain
characteristics of simultaneous surface and intra-atrial recordings of atrial fibrillation, Comput. Cardiol., vol. 21, pp. 781784, 1994.
[49] 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.
[50] J.J. Langberg, J.C. Burnette, and K.K. McTeague, Spectral analysis of the
electrocardiogram predicts recurrence of atrial fibrillation after cardioversion,
J. Electrocardiol., vol. 31, pp. 8084, 1998.

NOVEMBER/DECEMBER 2006