zhoujian cao et al- turbulence control with local pacing and its implication in cardiac...
TRANSCRIPT
-
8/3/2019 Zhoujian Cao et al- Turbulence control with local pacing and its implication in cardiac defibrillation
1/9
Turbulence control with local pacing and its implicationin cardiac defibrillation
Zhoujian Cao and Pengfei LiDepartment of Physics, Beijing Normal University, Beijing 100875, China
Hong ZhangZhejiang Institute of Modern Physics and Department of Physics, Zhejiang University,Hangzhou 310027, China
Fagen Xie Research Department, Kaiser Permanente, Pasadena, California 91101
Gang Hua
Department of Physics, Beijing Normal University, Beijing 100875, Chinaand BeijingHong KongSingapore Joint Center of Nonlinear and Complex Systems,Beijing Normal University Branch, Beijing, China
Received 19 October 2006; accepted 9 February 2007; published online 30 March 2007
In this review article, we describe turbulence control in excitable systems by using a local periodic
pacing method. The controllability conditions of turbulence suppression and the mechanisms un-
derlying these conditions are analyzed. The local pacing method is applied to control Winfree
turbulence WT and defect turbulence DT induced by spiral-wave breakup. It is shown that WTcan always be suppressed by local pacing if the pacing amplitude and frequency are properly
chosen. On the other hand, the pacing method can achieve suppression of DT induced by instabili-
ties associated with the motions of spiral tips while failing to suppress DT induced by the insta-
bilities of wave propagation far from tips. In the latter case, an auxiliary method of applying
gradient field is suggested to improve the control effects. The implication of this local pacing
method to realistic cardiac defibrillation is addressed. 2007 American Institute of Physics.
DOI: 10.1063/1.2713688
The only clinical method currently accepted in perform-
ing cardiac defibrillation is to apply large shocks to body
surface or directly to cardiac muscle.1,2
These large-
amplitude shocks may damage the cardiac tissue and
cause serious pains.3
Therefore, in both fields of nonlin-
ear science and cardiac physiology there are growing ex-
perimental and theoretical efforts to develop low-
amplitude defibrillation methods.49
In Ref. 7, the authors
performed suppression of Winfree turbulence by apply-
ing global periodic low-amplitude forcing.10
The difficulty
of injecting an external signal to whole cardiac tissue re-
stricts the practical utilization of this global method.
Therefore, so-called local and low-amplitude pacing
control,9
which applies low-amplitude signals to the local
boundary area of turbulent systems for global turbulence
suppression, may become a promising method for
defibrillating cardiac tissue. On one hand, boundary re-
gion control is very convenient, and on the other hand,
simple periodic pacing does not require instant measure-
ments and feedbacks of the system variables, and can be
easily realized in practical situations. In this article, we
review turbulence control in excitable systems by using a
local periodic pacing method, and both Winfree turbu-
lence (WT) and defect turbulence (DT) are
considered.11,12
It is shown that WT can always be sup-
pressed by local pacing as the pacing amplitude and fre-
quency are properly chosen. On the other hand, the pac-
ing method can achieve suppression of DT induced by
instabilities associated with the motions of spiral tips
while failing to suppress DT induced by the instability of
wave propagation far from tips. In the latter case, an
auxiliary method of applying gradient coupling is sug-
gested to improve the control effects. The implication of
this local pacing method to realistic cardiac defibrillation
is addressed.
I. INTRODUCTION
More and more experiments indicate that ventricular fi-
brillations VF are induced by transitions of cardiac electric
propagating waves from spiral waves to turbulent waves.13,14
Different mechanisms of turbulence formation in cardiac
systems are suggested, among which two cases will be dis-
cussed in this paper. One case of turbulence formation is due
to wave breakup through instabilities of two-dimensional
2D spiral waves. This spiral wave breakup turbulence is
associated with formation of topological defects, and thus is
called defect turbulence DT.6,15
In the other case, turbu-
lence is caused by Winfree instability, called Winfree turbu-
lence WT. Winfree turbulence, an unstable state of scroll
waves, is a special kind of spatiotemporal chaos that exists
exclusively in 3D excitable media, whereas its correspon-
dence in 2D system is stable spiral waves. WT results from
aAuthor to whom all correspondence should be addressed. Electronic mail:
CHAOS 17, 015107 2007
1054-1500/2007/171 /015107/9/$23.00 2007 American Institute of Physics17, 015107-1
http://dx.doi.org/10.1063/1.2713688http://dx.doi.org/10.1063/1.2713688http://dx.doi.org/10.1063/1.2713688http://dx.doi.org/10.1063/1.2713688http://dx.doi.org/10.1063/1.2713688http://dx.doi.org/10.1063/1.2713688http://dx.doi.org/10.1063/1.2713688 -
8/3/2019 Zhoujian Cao et al- Turbulence control with local pacing and its implication in cardiac defibrillation
2/9
the negative-tension instability of scroll waves, and is cur-
rently considered as one of the principal mechanisms of car-
diac fibrillation.7,16
Cardiac tissues are typical excitable me-
dia, and cardiac fibrillations are produced by various
dynamic transitions to turbulence. Therefore, defibrillation is
essentially a task of turbulence suppression in excitable me-
dia.
So far, a number of feedback and nonfeedback control
methods1723 have been suggested for suppressing spatiotem-poral chaos, among which nonfeedback approaches have
some advantages of flexibility and convenience in realistic
applications.9,22
In this paper, we are interested in one of the
nonfeedback methods, called the local pacing method, for
turbulence suppression in excitable media. For controlling
turbulence, we inject weak periodic signals just to a small
space region around the system surface. This local pacing
method is one of the low-amplitude defibrillation methods
developed recently in cardiac physiology. For instance, some
experiments explored the use of boundary periodic pacing
for cardiac defibrillation.2429
They found that the pacing had
only a local effect, and could not achieve global control offibrillation. Once the pacing was suspended, the local region
of capture was reinvaded by surrounding turbulent electrical
activity, and the tissue remains in a state of fibrillation. In
order to fully understand these failures and effectively im-
prove the technique of low-amplitude defibrillation, it is nec-
essary to study systematically local pacing methods for tur-
bulence suppression. In this review paper, we make a survey
on turbulence suppression with local pacing in excitable me-
dia. The control mechanism, efficiency, and the controllabil-
ity conditions are discussed in detail, and possible improve-
ment of the pacing method is discussed as well. Moreover,
the implication of this turbulence suppression to the problem
of cardiac defibrillation is also addressed. In the next section,
we introduce the models of excitable media used in this pa-
per for the analysis of turbulence suppression with local pac-
ing. In Sec. III and IV, we investigate WT and DT suppres-
sions, respectively. The influences of various parameters on
the control efficiency of the local pacing method are dis-
cussed in Sec. V. In the last section, we address some impli-
cation of this local pacing method to cardiac defibrillation.
II. MODELS, LOCAL PACING METHOD,AND GENERAL DISCUSSIONON CONTROLLABILITY
There are a number of mathematical models describing
the activities of the cardiac tissue, among which the Luo-
Rudy model
V
t=
1.0
CmINa + Isi + IK+ IK1 + IKp + Ib + D
2V 1
has been most extensively used. In Eq. 1, Vx , t is the
transmembrane potential of the cardiac cell at the space point
x, Is represent various iron currents through cell iron gates,
and Is are controlled by iron gate variables, which are con-
trolled in turn by the potential voltage V.9,30
Neglecting the complexity of various iron currents andgate variables, we can greatly simplify the dynamics by us-
ing only two variables,3133
the difference of electric poten-
tial and the total iron current. In the following, we use Bar-
kley and Br models,31,32
which have been extensively
investigated as typical excitable media for describing this
simplified dynamics. These two models together can be ex-
pressed by
u
t
=1
u1 u
u
v + b
a + 2u , 2a
v
t= fu v, 2b
where is the ratio of the time scale of the activator u over
that of the inhibitor v. For the Barkley model,31
function f
reads
fu = u 3
and for the Br model,32
the function f is specified as
fu =
0 0 u 1/3
1 6.75uu 12 1/3 u 1
1 1 u .
4
These two models have been used as the simplest homoge-
neous monodomain models representing the excitability of
cardiac electric activities. Although the two models are akin
to each other, there is an important difference between them.
Spiral waves in the Barkley model are always stable, i.e., the
spirals there never break up in 2D systems. This model is
suitable for the study of WT, which occurs in the 3D Barkley
model in certain parameter regions due to filament expan-
sion of the corresponding 3D scroll waves.7,34
As to the Br
model, its spirals have wealthy behaviors including 2D DTinduced by the Doppler effect,
31backfire instability,
32and so
on. This is therefore an ideal model studying DT and its
suppression. With local pacing, Eq. 2a becomes
u
t=
1
u1 uu v + b
a + 2u +xFcost,
x = 1 x0 x,
5
where is the boundary region to which we apply pacing.
We fix at the left site x =0 for 1D systems and left stripe
x =0 for 2D systems. The local pacing of Eq. 5 models the
low-amplitude control of cardiac systems through electric
poles placed on the surface of heart. Throughout the paper,
we apply no-flux boundary condition for different spacial
dimensions.
With local pacing control, we excite some local space
area periodically to generate regular propagating waves or
called controlling waves, and to use these waves to suppress
turbulence. The local control method functions in three key
processes of wave generation, wave propagation, and wave
competition. More specifically, the following three condi-
tions are of crucial importance for the success of the local
pacing method. a The periodic pacing must be able to gen-erate ordered controlling waves; b the controlling waves
015107-2 Cao et al. Chaos 17, 015107 2007
-
8/3/2019 Zhoujian Cao et al- Turbulence control with local pacing and its implication in cardiac defibrillation
3/9
must be able to propagate regularly in the medium; c the
propagating controlling waves must out-compete with turbu-
lent waves.
Now we analyze how these three conditions play roles in
turbulence suppression with local pacing. As to condition a,
we find that pacing frequency higher than certain threshold
cannot produce regular waves having the pacing frequency.
This condition can be investigated with a 1D chain. In Fig. 1,
we present a response curve of input-output frequencies ofthe Barkley model with a =1.1, b =0.19, =0.02, F=2.0.
When the pacing frequency increases too high 1.49, we
find that the output waves can no longer follow each excita-
tion pulse of the pacing, namely, the system output cannot
keep 1:1 frequency relation with the input, rather it can gen-
erate only 1: n n larger than or equal to 2 frequency re-
sponse. In Fig. 1 we observe 1:2 response for 2.88
1.61 and 1:3 response for 3.643.0, and so on. Note,
the highest output frequency is observed at the highest input
frequency of the 1:1 response region where the input fre-
quency is relatively low. And when the pacing frequency is
too high, it can only result in waves with low frequency see
1.26 in Fig. 1. Therefore, when the pacing frequency istoo high or too low, the output frequency cannot be high. The
former generates low output frequency due to1: n blocks
out=
nn2, while the latter generates low output fre-
quency out= due to low pacing frequency. In the discus-
sion of condition c, we will see that sufficiently high fre-
quency of controlling waves is a necessary condition for
turbulence suppression. In Figs. 2b and 2c, we show that
neither too high nor too low pacing frequencies can suppress
turbulence while suitable intermediate pacing frequency suc-
ceeds in doing so Fig. 2d. This is consistent with the
above analysis.
Condition b is self-evident, because in order to com-pete turbulent waves, the regular controlling waves must be
stable themselves. The problem for this condition is to illus-
trate the cases in which regular waves cannot propagate sta-
bly in excitable media. For instance, two mechanisms can
certainly make wave propagation unstable: one is the back-
firing effect and the other is alternans instability. In Fig. 3,
we compare essentially different behaviors of wave propaga-tions without a-c and with backfiring d-f. When
backfiring occurs, the forward-propagating waves produce
daughter waves propagating backward Fig. 3e, which
produce their daughters again. Finally, the whole space is
filled with these randomly distributed waves Fig. 3f. With
this backfiring instability, the controlling waves generated by
the periodic pacing are unstable, they produce turbulence by
themselves, and one can thus certainly not expect to use
these waves to suppress turbulence. The alternans phenom-
enon can be observed in Luo-Rudy35
and Aliev-Panfilov
models,33
which serve as typical models describing cardiac
dynamics. Originally, alternans means that a single cardiac
cell responds to external stimuli with long-short-long-short
period-doubling temporal behavior Fig. 4a.20,36
These
temporal alternans can produce also period-doubling in space
with long-short-long-short wavelengths shown in Fig. 4b.
If the alternans effect is too strong, successive waves can
collide with each other, leading to wave breakup and turbu-
lence Fig. 4c. If the controlling waves fall into this strong
alternans range, the above controllability condition b is not
met, and turbulence suppression with local pacing cannot be
achieved.
When the local pacing can generate regular controlling
waves, which can effectively propagate in the excitable me-
dium, there is still one more condition c that should befulfilled for successful turbulence suppression. Suppose two
FIG. 1. Output frequency versus input frequency for a driven 1D Barkley
model 5. The periodic pacing is applied at x = 0. a =1.1, b =0.19, =0.02,F=2.0. 1:1 frequency response is observed for 1.26. For 1.26, the
system cannot respond to each excitation of the pacing, and thus the con-
trolling waves generated by the pacing have frequency m : nnm to theinput frequency. The highest output frequency of the controlling waves,
which is most important for turbulence suppression, is determined by thelargest input frequency in the 1:1 region.
FIG. 2. Contour patterns of the 2D Br system a =1.1, b =0.19, =0.09, F
=2.0. a F=0. Turbulent state freely evolved from a spiral initial state.
b-d The boundary stripe control Eq. 5 is applied to the state a. b=1.34. Defect turbulence control with frequency that is too high no sup-pression. t=2000. c = 1.05. Defect turbulence control with frequencythat is too low no suppression. t=2000. d = 1.25. Successful defectturbulence suppression with suitable frequency. t=1000. Figure fromRef. 12.
015107-3 Local pacing control Chaos 17, 015107 2007
-
8/3/2019 Zhoujian Cao et al- Turbulence control with local pacing and its implication in cardiac defibrillation
4/9
waves wave A and wave B interact and compete with each
other. The phase-matching condition governing the motion ofthe interface between A and B reads
37
k1x + 1t+ = k2x + 2t+ , 6
where k1,2 are the wave numbers of waves A and B, respec-
tively, and x is the distance between the interface point and
some reference point e.g., wave source. 1,2 are the fre-
quencies of two wave chains and is the phase angle of the
interface point. Thus the speed of the interface VIdx1
dt=
dx2
dtis given by
VI =1 2
k1 + k2. 7
Positive negative velocity VI represents that wave A wave
B suppresses wave B wave A. This formula is consistent
with the finding that in excitable media, propagating waves
with the highest excitation frequency can always overtake all
other waves with lower frequencies.38
In conclusion, in order
for the controlling waves generated by the local pacing tosuppress turbulent waves in excitable media, the frequency
of the former must be higher than that of the latter.
III. WINFREE TURBULENCE SUPPRESSIONWITH LOCAL PACING
From the analysis of Sec. II, one can conclude that all
stable spiral waves can be suppressed by local pacing with
amplitude and frequency properly chosen. This strong con-
clusion can be understood heuristically. A spiral wave tip
serves as a pacing source generating wave. In the presence of
stable spiral waves, local pacing with the spiral frequencycan certainly satisfy the above conditions a and b since
the spiral wave tip can successfully generate and propagate
stable spiral waves. Generically, the frequency of the spiral
denoted by s is not the highest frequency say, h sup-
ported by the medium. We can then easily find a pacing
frequency sh with which the local pacing can still
satisfy conditions a and b on one hand, and the frequency
of the controlling waves generated by this pacing is higher
and that of the spiral and then can win the competition with
the stable spiral on the other hand since s, and condi-
tion c is satisfied. WT is induced by the 3D effect of stable
2D spiral waves.7
Due to the negative tension of the filament
of the corresponding 3D scroll waves, all the above reason-
ing on the controllability of stable spirals exists in the pro-
cess of WT suppression note, filaments that cause instability
of scroll waves to WT do not exist in controlling waves.
This makes us propose a conjecture that all WT can be sup-
pressed by local pacing with proper pacing amplitude and
frequency. We use the Barkley model Eqs. 1 and 2 to
discuss the problem of WT suppression. It is well known that
all spirals in the 2D Barkley model are stable, and in certain
parameter regions scroll waves of the 3D Barkley model can
be destabilized via filament expansion instability,7
develop-
ing WT as shown in Fig. 5. Now we apply periodic pacing to
a small cube nnn sites region on a surface11
and adoptthe pacing protocol suggested in Ref. 7,
FIG. 3. Comparison of wave propagation with and
without backfiring. ac Wave propagation without
backfiring. df Wave propagation with backfiring.a and d, b and e, c and f are at the same times,respectively.
FIG. 4. Schematic figures of alternans and alternans DT in the Luo-Rudy
model. a and b Alternans effect of long-short-long-short period-doublingmotion in the 1D Luo-Rudy model. a An alternans evolution of the actionpotential V at an arbitrary space point. b An evolution contour pattern in
the x-t plane. c A snapshot contour pattern of alternans DT in the 2DLuo-Rudy model at an arbitrary moment.
015107-4 Cao et al. Chaos 17, 015107 2007
-
8/3/2019 Zhoujian Cao et al- Turbulence control with local pacing and its implication in cardiac defibrillation
5/9
F=1
au1 ubf. 8
Figure 6 demonstrates the process of WT suppression by
local pacing. The local excitation on a small cube stimulates
spherical target waves, which propagate in the excitable me-
dium, invading the turbulent surrounding, and suppress tur-
bulence by pushing filaments of scroll waves out of the ex-
citable medium during their propagation Figs. 6a6d. In
Figs. 6e and 6f, we show how the ordered waves wipe out
the remaining filaments and drive the system to the desired
rest state, after lifting the periodic forcing at the turbulent
state of Fig. 6c.
In Fig. 7, we study a quantity important for the control
efficiency, the control speed that is defined as R = 1 /T with Tbeing the time needed for turbulence suppression. R depends
on forcing frequency, injection area, and other pacing param-
eters. Figure 7 shows the parameter range where the local
pacing can successfully suppress the existing turbulent scroll
waves. There are several characteristic boundaries to this
range corresponding to the controllability conditions ac
emphasized in Sec. II. First, this range is restricted within a
frequency zone Fig. 7a, and frequencies above or below
this zone do not provide successful turbulence suppression.
The forcing frequency needs to be higher than a threshold 00 1.19 is the rotation frequency in unforced 2D system
due to condition c. This is because the target waves gener-ated by the local excitation can suppress existing scroll
waves only if its frequency is higher than that of the latter.38
There is an upper frequency bound for effective control. The
target waves are generated by the periodic signals in the
presence of turbulence and scroll waves. In order to effec-
tively stimulate target waves from the existing scroll wave
background, the forcing frequency should not be too far from
0 for achieving 1:1 resonant excitation due to condition
a. This gives rise to the upper bound of the zone of Fig.
7a. Secondly, the injection area should be sufficiently large.
Our WT suppression with cubic stimulation fails for n4
Fig. 7
b
. The reason for this condition is clear. In order toproduce regular waves, the injection area needs to be larger
than a threshold so that the wave front generated has curva-
ture smaller than a critical value for allowing target wave
propagation otherwise, no spherical wave front can be
formed, condition a.39 Nevertheless, when n is above the
threshold, the speed of turbulence annihilation is no longer
sensitive to n. In other words, further increase of n does not
further reduce the suppression time. This characteristic al-
lows us to take some optimal control area for performing WT
suppression. In Fig. 8, we test our conjecture of WT control-
lability with local pacing by examining the 3D Barkley
model for different parameter values. With =0.02, b =0.19
fixed, we vary a in the interval a = 0.98,1.16 where WT
occurs. Note, WT does not exist for smaller and larger a. For
a0.98, the system cannot support any waves due to too
low excitability,7
and the system has only a single stable
state, the stationary homogeneous rest state. With a1.16,
scroll waves become stable because of positive tension of
filaments.7
Figure 8 satisfactorily confirms at least, for the
Barkley model the conjecture that WT can be always sup-
pressed by local pacing.
As a possible mechanism of cardiac fibrillation,7,16
WT
suppression is one of the crucial tasks in cardiac defibrilla-
tion. Since local pacing methods are effective in WT sup-
pression in excitable media, they have some interesting im-plication in defibrillation. As a possible defibrillation
FIG. 5. Evolution of WT in the 3D Barkley model. a =1.1, b =0.19, and
=0.02. a Initial state with a regular filament circle located at the center ofsimulation domain. b Filament expansion to an irregular while still con-tinuous curve at t=50. c The expanded filament touches space boundaries,and the boundaries cut the filament into many segments. t=150. d A suf-ficiently developed WT state at t=250.
FIG. 6. WT control with local pacing on a 555 cube in the center of the
back surface of the space. bf=0.75,=1.6 and all other parameters are the
same as Fig. 5. The pacing is applied to Fig. 5d at t=250. ad The localpacing generates spherical target waves near the injected area, which then
propagate in the space and push turbulence and all scroll wave filaments out
of the excitable medium. e,f The evolution of the medium after the in- jection signals are lifted at t=331 i.e., from the state c. The systemapproaches the desired rest state without the external forcing, and complete
black indicating the rest state is observed after t360. Figure fromRef. 11.
015107-5 Local pacing control Chaos 17, 015107 2007
-
8/3/2019 Zhoujian Cao et al- Turbulence control with local pacing and its implication in cardiac defibrillation
6/9
method, the present local stimulation shows a number of
advantages in comparison with some other low-amplitude
defibrillation methods, such as global controlling methods.
First, the global method7
needs to inject external signals to
all space sites, and this is inconvenient in many practical
situations e.g., it is difficult to apply electrical impulses to a
large area of interior cardiac body. On the contrary, it is
much more convenient for local pacing methods to injectsignals to partial areas around the surface of the system.
Second, the turbulence suppression speed of our local stimu-
lations is faster11
than that of the global ones in Ref. 7 if
similar powers of pacing signals are applied in both cases. It
is worthwhile to remark that in Fig. 6 we suppress WT by
applying signals to 53 sites only among the total 1503 sites
with signal amplitude being of the same order as the system
variables. This is desired for developing a low-amplitude
defibrillation method.
IV. DEFECT TURBULENCE SUPPRESSION
WITH LOCAL PACING
Considering another possible mechanism of cardiac fi-
brillation, we study DT suppression in excitable media. Ap-
parently, the previous arguments on the general controllabil-
ity of stable spirals and WT are no longer valid for DT
control because spirals are now not stable. In excitable me-
dia, spirals can break up through different mechanisms. They
may break up in the regions near the spiral tips, such as
through the Doppler effect,31
or may break up through the
instabilities in arm wave propagation far from the tips, such
as through backfiring32
and alternans instabilities.20,36
We
will observe in this section that the former DT can be easily
suppressed by local pacing while the latter cannot be doneso, and the reasons for these observations will be explained
later. Let us consider the 2D Br model with a =1.1, b
=0.19, and we apply periodic pacing of Eq. 5 with F=2.0
to analyze the problem of controllability of DT by external
pacing, which is applied to the left boundary of the space,
x 0 ,x ,y0 ,L. In Fig. 9, the controllable region on
the - plane is presented. Without control, the entire region
0.069 corresponds to breakup spiral to DT in this model,
and Doppler-effect-induced DT is justified in the interval
0.0690.103. It is interesting to find that, with the
boundary pacing, the controllable region covers all the Dop-
pler DT domain, and this region is ended at =0.103, whichis just the threshold of backfiring.
32The coincidence of the
controllable parameter region boundary with the backfiring
boundary confirms our analysis of controllability condition
FIG. 7. Turbulence suppression speed
R = 1 /T of the cubic nnn injec-tion plotted for different forcing pa-
rameters where T is the time needed
for turbulence suppression. a n = 6,bf=0.75. R plotted vs . b bf=0.75,
=1.6. R vs n. Figure from Ref. 11.
FIG. 8. Controllable region of WT suppression in the Barkley model. With
=0.02, b =0.19, WT exists only in the parameter region 0.98a1.16.
Thus WT in all this range can be suppressed by boundary local pacing withthe pacing amplitude and frequency properly chosen.
FIG. 9. Controllability region, where all spiral waves and DT wavelets can
be suppressed, in the 2D Br model. The system state is DT for 0.069.
=0.103 is the critical parameter above which backfiring takes place. The
upper line and the lower line are the highest and the lowest frequencies of
controllability. Pacing with frequency between these two curves can sup-press spirals and DT successfully.
015107-6 Cao et al. Chaos 17, 015107 2007
-
8/3/2019 Zhoujian Cao et al- Turbulence control with local pacing and its implication in cardiac defibrillation
7/9
b in Sec. II. In the Doppler DT region, DT is induced by
collisions between spiral tips topological defects and spiral
wave arms.32
Since the controlling waves do not have topo-
logical defect see Fig. 2d and they do not suffer from the
Doppler instability, the controlling waves can surely suppress
Doppler DT when the local pacing frequency is chosen larger
than that of the spiral wavelets. On the other hand, backfiring
makes propagation of any regular waves fail, and in the pres-
ence of such instability the controlling waves generated bythe local pacing must fail in suppressing DT no matter how
one adjusts the pacing amplitude and frequency, because DT
can be induced by the controlling waves themselves during
their propagation. The parameter regions with these kinds of
instabilities are thus called an absolutely uncontrollable re-
gion for local pacing. In Fig. 9 the backfiring region,
0.103, is a kind of absolutely uncontrollable region. The
existence of an absolutely uncontrollable region may be an
interpretation of failures of some experiments for cardiac
defibrillation by applying local periodic pacing.2429
It is em-
phasized that backfiring is a sufficient but not a necessary
condition for an absolutely uncontrollable parameter region.Other instabilities occurring in wave propagation, e.g., wave-
breakup induced by alternans, can also determine the abso-
lutely uncontrollable parameter domain.
It is possible that some cardiac fibrillation may appear
right in the parameter domain of absolute uncontrollability; it
is interesting to seek realistic means to remedy the condition
b for improving the local pacing method. In the following,
we describe a convenient method to aid local pacing to con-
trol backfiring DT. Our method is to introduce an additional
gradient field term into Eq. 5,4043
u
t =
1
u1 uu v + b
a + 2u +xFcos t E
u
x. 9
It is argued that in some physical situations, the gradient
term in Eq. 9 can be easily realized. For instance, if the
variables of the system represent some densities of irons, the
gradient term appears when we apply a certain electric po-
tential difference on the left and right boundaries. In this
case, the gradient term is realized also conveniently by
boundary actions. In Fig. 10a, we compare the DT control-
lable regions of Eq. 9 in the - plane with different gra-
dient intensities. It is shown that the gradient term improves
the control effect surprisingly well. The physical mechanism
underlying this striking improvement can be clearly under-
stood. The essential influence of the gradient force to the
wave propagation in excitable media is to make defects and
waves move along the gradient direction Fig. 10b. This
influence has two effects simultaneously. On one hand, it
reduces the frequency of spiral and turbulent waves of the
system due to the Doppler effect Fig. 10c. On the other
hand, it makes the pacing-generated ordered waves leave the
pacing boundary faster along the gradient direction, and this
allows the system to keep 1 : 1 frequency response against
faster pacing and to generate controlling waves with largerfrequency. The former effectively moves down the low con-
trollable frequency threshold while the latter considerably
moves up the high-frequency boundary of the controllable
region. These two tendencies together effectively push back-
firing threshold B to larger Fig. 10d, and makes the
large absolutely uncontrollable parameter area controllable.From Fig. 10 it is obvious that the gradient coupling can
greatly improve DT control with local pacing. Nevertheless,
the applicability of this gradient term to realistic cardiac
defibrillation is still an open problem. First, it is not clear
how to apply the gradient coupling in cardiac systems. More-
over, it is commonly believed that cardiac fibrillation is, to a
large extent, related to DT induced by alternans
instability,20,36
and the effect of gradient force on suppression
of alternans DT is another open problem.
V. OTHER FACTORS INFLUENCING
THE EFFECTS OF LOCAL PACING
Besides the pacing frequency and pacing area discussed
in previous sections, there are some other factors influencing
the effects of the local pacing method. First, the pacing am-
plitude must be a relevant factor. It is trivial to conclude that
a pacing cannot stimulate any controlling waves and thus
cannot achieve any turbulence suppression if the pacing am-
plitude is too small. There exists a threshold for the ampli-
tude, and the periodic local pacing can generate controlling
waves only if the pacing amplitude is larger than this critical
value. Another less trivial effect of pacing amplitude is that
changing amplitude can change the 1:1 response region of
Fig. 1, and can indirectly change the controllable parameterregion. We again use the Br model to illustrate this effect. In
FIG. 10. Numerical results of Eq. 9 with F= 2. a The controllable zonewithout circles, E= 0 and with disks, E=0.35; stars, E=0.6 gradient field.The backfiring thresholds B for different Es are shown in the figure with
vertical lines. b The velocity of pacing-generated waves plotted vs E for=0.13. c The principal frequency of turbulent waves plotted vs E for =0.13. d B vs E for Eq. 9. Figure from Ref. 12.
015107-7 Local pacing control Chaos 17, 015107 2007
-
8/3/2019 Zhoujian Cao et al- Turbulence control with local pacing and its implication in cardiac defibrillation
8/9
Fig. 11a, we do the same as Fig. 1 with different pacing
amplitudes applied. It is interesting to see that increasing the
pacing amplitude can considerably increase the highest fre-
quency of the controlling waves. In Fig. 11b, the highest
response frequency is plotted against pacing amplitude. The
highest response frequency increases with amplitude for
small F, and saturates to a certain value for large F. With this
picture we can understand why increasing the pacing ampli-
tude can effectively improve the pacing results for relativelysmall F, while as the amplitude is sufficiently large, a further
increase of F no longer changes the control effect anymore.
In all the previous discussions, we used only a cosine
signal for the local pacing. It has been found that a different
form of pacing signal can also considerably change the pac-
ing results. This is because that different pacing form can
make the highest frequency of the controlling waves differ-
ent. And this again influences the effect of turbulence sup-
pression indirectly. This problem has been discussed in Ref.
9 in detail.
One more relevant factor of turbulence suppression with
local pacing is the system size. Because some instabilities,
such as alternans and backfiring instabilities, emerge only for
sufficiently large systems, changing the system size may also
essentially change the control results. As an example, we
compute the input-output frequency response curves for the
1D Luo-Rudy model and adopt the local pacing strategy of
Ref. 9. It is shown that the response behavior for small sys-
tem size Fig. 12a, and also Fig. 1 in Ref. 9 is considerably
different from that for large system size Fig. 12b. From
these response behaviors, one concludes that turbulence sup-
pression with local pacing in small systems is easier than in
large systems, a conclusion familiar to experimentalists in
this field.
There are certainly some other factors influencing the
effects of local pacing. To our understanding, the factor of
input-output frequency relation is the most important. The
effects of many other factors can often influence the controlresults by modifying the input-output frequency relations,
similar to what is shown in Figs. 11 and 12.
VI. CONCLUSION
In this paper, we present a survey of local pacing control
in suppressing turbulence in excitable media. We show that
the local pacing method can effectively suppress WT and DT
induced by near-field instabilities associated with spiral
tips. The local pacing method fails in DT suppression when
the turbulence occurs via some instabilities in wave propa-
gation, such as backfiring and alternans instabilities. In the
latter case, some auxiliary method of gradient coupling is
suggested to aid the local pacing to perform DT suppression.
These understandings are expected to be useful in develop-
ing practical methods for cardiac defibrillation. It was argued
that WT may be one of the major causes of fibrillation.16,44
The above controllability prediction of WT may be of help in
searching low-amplitude local pacing methods to perform
FIG. 11. a The same as Fig. 1 with different pacingamplitudes applied. a =1.1, b =0.19,=0.03. It is ob-
served that different pacing amplitudes may result in
different input-output relations. b The highest outputfrequency vs pacing amplitude. When F2.2, the high-
est output frequency saturates to a certain constant.
FIG. 12. The same as Fig. 1 with the 1D Luo-Rudy model computed. The system dynamics and parameters are given in Ref. 9 and the response curves areobtained for different system sizes. a L =0.84 cm. Figure from Ref. 9. b L =8.4 cm.
015107-8 Cao et al. Chaos 17, 015107 2007
-
8/3/2019 Zhoujian Cao et al- Turbulence control with local pacing and its implication in cardiac defibrillation
9/9
defibrillation in this case. On the other hand, propagating
waves breaking up to DT often occur for steeply sloped res-
titution curves,45,46
where low-amplitude pacing may not
succeed in cardiac defibrillation. We suggest that in these
cases one has to search some effective auxiliary methods
cooperating with the low-amplitude pacing to perform car-
diac defibrillation. The methods reviewed here focus on WT
and DT suppression only in some simple models of excitable
media. In the future, we will investigate these methods inmodels representing more realistic electric and chemical ac-
tivities of heart systems, and develop applications of these
methods in direct cardiac defibrillation experiments.
1M. S. Eisenberg et al., Sci. Am. 2545, 25 1986.
2L. W. Piller, Electronic Instrumentation Theory of Cardiac Technology
Staples Press, London, 1970.3L. Tung, Proc. IEEE 84, 366 1996.
4V. N. Biktashev and A. V. Holden, J. Theor. Biol. 169, 101 1994.
5A. V. Panfilov et al., Phys. Rev. E 61, 4644 2000.
6S. Sinha, A. Pande, and R. Pandit, Phys. Rev. Lett. 86, 3678 2001.
7S. Alonso, F. Sagues, and A. S. Mikhailov, Science 299, 1722 2003.
8M. Vinson et al., Nature London 386, 477 1997.
9
A. T. Stamp, G. V. Osipov, and J. J. Colins, Chaos 12, 931 2002.10Y. Braiman and I. Goldhirsch, Phys. Rev. Lett. 66, 2545 1991.11
H. Zhang, Z. Cao, N. Wu, H. Ying, and G. Hu, Phys. Rev. Lett. 94,
188301 2005.12
Z. Cao, H. Zhang, F. Xie, and G. Hu, Europhys. Lett. 75, 875 2006.13
J. M. Davidenco, A. V. Pertsov, R. Salomonsz, W. Baxter, and J. Jalife
Nature London 355, 349 1992.14
F. X. Witkowski, Nature London 392, 78 1998.15
P. Coullet, L. Gil, and J. Lega, Phys. Rev. Lett. 62, 1619 1989.16
A. T. Winfree, Science 266, 1003 1994.17
G. Hu and Z. Qu, Phys. Rev. Lett. 72, 68 1994.18
I. Aranson, H. Levine, and L. Tsimring, Phys. Rev. Lett. 72, 2561 1994.19
V. S. Zykov et al., Phys. Rev. Lett. 92, 018304 2004.20
G. M. Hall and D. J. Gauthier, Phys. Rev. Lett. 88, 198102 2002.21
G. Baier, S. Sahle, and J. Chen, J. Chem. Phys. 110, 3251 1999.22
H. Zhang, B. Hu, and G. Hu, Phys. Rev. E 68, 026134 2003.
23H. Zhang et al., Phys. Rev. E 66, 046303 2002.
24M. Allessie, C. Kirchhof, G. J. Scheffer, F. Chorro, and J. Brugada, Cir-
culation 84, 1689 1991.25
C. Kirchhof, F. Chorro, G. J. Scheffer, J. Brugada, K. Konings, Z. Zete-
laki, and M. Allessie, Circulation 88, 736 1993.26
B. H. Kenknight, P. V. Bayly, R. J. Gerstle, D. L. Rollins, P. D. Wolf, W.
M. Smith, and R. E. Ideker, Circ. Res. 77, 849 1995.27
E. G. Daoud, B. Pariseau, M. Niebauer, F. Bogun, R. Goyal, M. Harvey,
K. C. Man, S. A. Strickberger, and F. Morady, Circulation 94, 1036
1996.28J. M. Kalman, J. E. Olgin, M. R. Karch, and M. D. Lesh, J. Cardiovasc.
Electrophysiol. 7, 867 1996.29
A. Cappucci, F. Ravelli, G. Nollo, A. S. Montenero, M. Biffi, and G. Q.
Villani, J. Cardiovasc. Electrophysiol. 10, 319 1999.30
J. Keener and J. Sneyd, Mathematical Physiology Springer-Verlag, 1998.31
D. Barkley, M. Kenss, and L. S. Tuckerman, Phys. Rev. A 42, 2489
1990.32
M. Br and M. Eiswirth, Phys. Rev. E 48, R1635 1993; M. Or-Guil, J.
Krishnan, I. G. Kevrekidis, and M. Br, Phys. Rev. E 64, 046212 2001.33
R. R. Aliev and A. V. Panfilov, Chaos, Solitons Fractals 7, 293 1996; A.
V. Panfilov, Phys. Rev. Lett. 88, 118101 2002.34
V. N. Biktashev, A. V. Holden, and H. Zhang, Philos. Trans. R. Soc.
London, Ser. A 347, 611 1994; F. H. Fenton et al., Chaos 12, 852
2002.35
C. H. Luo and Y. Rudy, Circ. Res. 68, 1501 1991.
36G. M. Hall and D. J. Gauthier, Phys. Rev. Lett. 88, 198102 2002.37
K. Nam, E. Ott, M. Gabbay, and P. Guzdar, Physica D 118, 69 1998.38
F. Xie, Z. Qu, J. N. Weiss, and A. Garfinkel, Phys. Rev. E 59, 2203
1999.39
V. S. Zykov, Simulation of Wave Processes in Excitable Media Manches-
ter University Press, New York, 1987, p. 117.40
O. Steinbock, J. Schutze, and S. C. Muller, Phys. Rev. Lett. 68, 248
1992.41
A. P. Munuzuri et al., Phys. Rev. E 48, R3232 1993.42
V. I. Krinsky, E. Hamm, and V. Voignier, Phys. Rev. Lett. 76, 3854
1996.43
M. Gabbay, E. Ott, and P. N. Guzdar, Phys. Rev. E 59, 2443 1999.44
R. Gray et al., Science 270, 1222 1995.45
A. Karma, Chaos 4, 461 1994.46
A. Garfinkel et al., Proc. Natl. Acad. Sci. U.S.A. 97, 6061 2000.
015107-9 Local pacing control Chaos 17, 015107 2007