Biomedical Signal Processing and Control 4 (2009) 57–66

Contents lists available at ScienceDirect

Biomedical Signal Processing and Control

journal homepage: www.elsevier.com/locate/bspc

Synchronization of two coupled pacemaker cells based on the

phase response curve
Hossein Gholizade-Narm a, Asad Azemi b, Morteza Khademi a, Masoud Karimi-Ghartemani c,*
a
Engineering Faculty, Ferdowsi University of Mashhad, Mashhad, Iran
b
College of Engineering, Pennsylvania State University, Delaware County Campus, USA
c
Electrical Engineering Faculty, Sharif University of Technology, Tehran, Iran

A R T I C L E I N F O A B S T R A C T

Article history: In this paper, the synchronization of a pair of pacemaker cells as Sino-Atrial (SA) and Atrio-Ventricullar
Received 18 December 2007 (AV) nodes have been studied and a new approach for synchronization, based on the concept of Phase
Received in revised form 2 November 2008 Response Curve (PRC), has been proposed. The paper starts with presenting the necessary and sufficient
Accepted 6 November 2008
conditions for synchronization in terms of the PRC parameters. Such conditions are time dependent and
thus, the paper proceeds with deriving some sufficient conditions, which are not time dependent. The
Keywords: time-delay between the firing time of SA node and when it reaches the AV node is also considered. When
Synchronization
the conditions for spontaneous synchronization are not valid, the synchronization is achieved by
Pacemaker
applying pulses to the AV or the SA nodes or to both of the nodes, depending on the accessibility. The
PRC
Stimulus subject has been investigated and sufficient conditions were achieved for all three cases. In each case, the
Pulse dynamical equations of coupled pacemakers have been determined and the stability analyses of delay
dynamical equations between discharges of two pacemakers were performed. The number of excitation
pulses and the time intervals for applying them to accessible pacemaker(s) were obtained and eventually
some numerical examples were simulated to approve the accuracy of the theoretical results and
conditions.
ß 2008 Elsevier Ltd. All rights reserved.

1. Introduction
The sinus node is the origin of cardiac activity and generates the
contraction commands but the mechanism by which a sinus
rhythm is determined, is not fully understood [1]. It has been
admitted especially by physicians, that normally a cell having the
highest frequency drives the other cells [2–4]. However, Dehaan
and Hirakow [5] reported that on the contact of pairs of
spontaneously beating myocytes of chicken hearts, some were
synchronized at a higher frequency than that of the higher
member, some at a frequency lower than the lower member, and
the others at an intermediate frequency of the two. Sano et al. [6]
found an inter-neuronal electronic influence on their firing rates in
the sinus node of a rabbit and obtained the phase response curve
(PRC).
These results, suggest that the synchronization among pace-
maker neurons plays a crucial role in the determination of the
sinus rhythm and that the mechanism appears to be complex.
Electrophysiological studies have suggested that the activity of
* Corresponding author.
E-mail address: karimig@shariif.edu (M. Karimi-Ghartemani).
cardiac cells with automaticity (e.g. in the Sino-Atrial (SA) node,
Purkinje network, Atrial and Ventricular myocardium) can be
modulated by stimulating current pulses (sub-threshold depolar-
izing or hyperpolarizing) applied extra-cellularly [6–8]. The effects
of external stimuli on the frequency of biological oscillators were
observed in a wide range of species, and their overall character-
istics can be well described by a PRC [7–9]. Systematic perturba-
tion techniques (by stimulating the pacemaker at various phases of
its intrinsic cycle) have been applied to experimentally investigate
and to establish the shape of the PRC [10,11].
Numerous modeling has been performed on cardiac cells and
electric propagation based on PRC until now. Wei et al. [12]
modeled cardiac in three dimensional form based on the PRC and
simulated supraventricular tachycardias and the Wolff–Parkinson–
White (WPW) syndrome. In [13], a cardiac computer model is used
with M cells and three dimensional excitation sequences during
torsade de point are detected. The initiation, propagation and
termination of that arrhythmia are also simulated in [13]. The effects
of automatic Vagal and sympathetic nervous system were
investigated by Abramovitch and Akselrod in 1998 [14] based on
the PRC. Moreover, the effects of a train of pulses on pacemakers
were studied and the entrainment regions of the pacemaker with
input pulses by the ratio m:n were obtained also. In [15], atrium has

1746-8094/$ – see front matter ß 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bspc.2008.11.001
58 H. Gholizade-Narm et al. / Biomedical Signal Processing and Control 4 (2009) 57–66
been simulated by a two-dimensional array of pacemaker cells,
composed of a region of SA node pacemaker cells surrounded by
atrial pacemaker cells. Then, the atrium normal activity, transition
from normal to various arrhythmias and atrial fibrillation was
studied. In [16], the SA node is modeled as an array and procedure
of impulse generation and was investigated based on the PRC.
Tsalikakis et al. [17] discovered the SA node dynamical response to
an external stimulus with various amplitudes and durations on
the cell model of [18] and obtained phase transition curves and
PRCs.
There is an important advantage of working with the PRC for
modeling and simulation and that is disengaging differential
equations. It is enough to know only the intrinsic cycle length, that
identifies the beat rate of cell without external stimulus, and the
PRC parameters, that identify the interaction intensity with other
cells near it. This is true if the stimulus amplitude is assumed
constant.
This paper treats the synchronization subject from the point of
view of dynamical equations and stability analysis. Moreover, the
paper investigates the ways of synchronizing the two cells. The
spontaneous synchronization conditions of a set of two coupled
pacemakers are determined in terms of novel time-independent
conditions. Moreover, when those conditions are not satisfied, the
number of stimulus pulses and the instances in which these pulses
should be applied are obtained to ensure synchronization of the
two pacemaker cells. Various scenarios were studied depending on
the accessibility of the nodes and simulations are performed to
confirm the results.
This paper is organized as follows. In Section 2, the PRC is
introduced and the problem statement is presented in Section 3.
The spontaneous synchronization conditions are investigated in
Section 4 and then the three cases of having access to only SA node,
Atrio-Ventricullar (AV) node and to both SA and AV nodes are
studied in Section 5. Section 6 presents simulation results, which
confirm the theoretical results, obtained in the previous parts. In
Section 7, the conclusion of this paper is expressed.
2. The PRC
A phase response curve tabulates the transient change in the
cycle period of an oscillator induced by a perturbation as a function
of the phase at which it is received. The phase response curve in
cardiac literature is referred to a curve which defines the phase
shift (Dw) of the discharge of a pacemaker cell with a constant
intrinsic cycle length as a function of the phase at which an
external stimulus is applied to the pacemaker cell [14]. The PRC, in
cardiac literature, is a mathematical definition for phase resetting
of cardiac pacemakers [19,20]. Assume t is intrinsic cycle length of
pacemaker (ms), T external depolarization cycle length (ms), d, a
and w are unit-less parameters belonging to [0,1] interval and
representing the delay, acceleration and the crossover point of the
PRC respectively, d, coupling interval, namely the time from
previous beat of the pacemaker cell to the time of the stimulus
(ms). Fig. 1 symbolically depicts the above description.
Fig. 1. Intrinsic cycle train pulses and external stimulus.
Fig. 2. A modeled phase response curve (PRC), described by three parameters: d,
delay; a, acceleration; and w, the crossover point. It shows the phase shift Dw of
pacemaker cells as a function of the phase w at which an external stimulus is applied
to the pacemaker cell.
The phase (w) at which the stimulus is applied is
d
’¼ (1)
t
That belongs to interval [0,1]. The phase shift of the next cycle, Dw,
is determined from the PRC (here the PRC is taken piecewise linear
for simplification):

ðd=wÞ  ’ ð0  ’ < wÞ
PRC  D’ ¼ (2)
½a=ð1  wÞ  ð’  1Þ ðw  ’  1Þ
t new ¼ t intrinsic þ D’  t intrinsic (3)
fnew ¼ gð’old Þ ¼ ’old  D’ (4)
where wnew is new phase, wold old phase, Dw phase shift, tnew new
cycle length and tintrinsic, intrinsic cycle length. A modeled PRC is
shown in Fig. 2.
Here we take PRC as piecewise linear for simplicity (similar to
[14–16]). However, in the absence of such a simplified situation,
similar analysis can be performed to address the topic for a
nonlinear model, which we have avoided here.
3. The problem statement
There is a pair of pacemaker cells as SA and AV nodes (Fig. 3).
Assume SA node is defined with T1 as its intrinsic cycle length and
PRCSA ¼ ½dSA aSA wSA  which defines the external stimulus effects and
AV node with T2 as its intrinsic cycle length and PRC ¼ ½d a w which
describes SA node effects and PRCAV ¼ ½dAV aAV wAV  which deter-
mines external stimulus effects. Assuming T2 > T1 (which is often the
case), we would like to obtain the desired time-independent
sufficient conditions for 1:1 synchronization of the pacemakers.
When two coupled oscillators work, depending on the coupling
strength, they can oscillate synchronously or asynchronously. The
coupling strength is specified by PRC’s parameters, d and a. The
smaller the parameters, the weaker the coupling is. Let us first find
some sufficient conditions for spontaneous synchronization. If the
two pacemakers do not hold the conditions, then depending on
where we have access, we will propose a method and an algorithm
to find the number of pulses (stimuli) and their corresponding time
instances to achieve synchronization.
4. Spontaneous synchronization
Suppose two pacemakers synchronize in the sense of frequency
without applying external stimulus (1:1 synchronization). Since in
general, the intrinsic beat rate of the SA node is higher than that of
the AV node one, therefore the discharge time of SA node should be
 H. Gholizade-Narm et al. / Biomedical Signal Processing and Control 4 (2009) 57–66 59

Fig. 3. The SA and AV pacemakers with characteristics T1, PRCSA and T2, PRC, PRCAV, respectively.

If two pacemakers synchronize, d(.) must converge to a positive

fixed point. The fixed point is obtained as follows:

ð1  ða=1  wÞÞDT
dðnÞ ¼ dðn  1Þ ¼ dF ) dF ¼
ða=1  wÞ
(11)
dF T1  d
F
’F1 ¼ ; ’F2 ¼
T1 T2
Fig. 4. Discharge times of SA and AV nodes.
Notice that (10) represents a discreet-time dynamical system
in a period that accelerates the AV node discharge time (Fig. 4). This with degree one. Therefore, the absolute of its eigenvalue must be
AV advancing continues until they are synchronized at a fixed less than unity for asymptotic stability as shown below.
delay (dF) between the SA and AV nodes and the AV node beat rate
a
reaches the SA node beat rate. jA þ 1j < 1 )  2 < A < 0; A¼ (12)
1w
The dynamical equations of discharge time of SA and AV nodes
and delay time between them are Since AV node should be discharged after SA node, therefore 0 < d(.)
must hold. According to (12), ða=1  wÞ > 0 and for 0 < dF we must
t 1 ðnÞ ¼ t 1 ðn  1Þ þ T 1
(5) have 1 > ða=1  wÞ, therefore the condition (12) changes as
t 2 ðnÞ ¼ t 2 ðn  1Þ þ T 2 þ D’2 ðn  1ÞT 2
a
and 0< <1 (13)
1w
t 2 ðn  1Þ  t 1 ðn  1Þ According to (10) and condition (13), if 0 < d(0) < dF then d(.)
’1 ðn  1Þ ¼ ; ’2 ðn  1Þ
T1 becomes increasing and its maximum is dF. Consequently for
t 1 ðnÞ  t 2 ðn  1Þ 0 < d(.) it is only enough that 0 < d(0) < dF.
¼ (6)
T2 On the other hand, using (9), from condition (8) and
a substituting from (6), we get:
D’1 ð:Þ ¼ 0; D’2 ðn  1Þ ¼ ð’ ðn  1Þ  1Þ (7)
1w 2
T 1  dðn  1Þ
where t1(.), t2(.) are discharge times of the SA and AV nodes w< <1 (14)
T2
respectively, w1 is the phase that determines the effect of AV node
on SA node (if bidirectional coupling is considered instead of A sufficient condition for (14) is wT 2 < minn ðT 1  dðn  1ÞÞ and
unidirectional) and w2 is the phase that determines the effect of SA maxn ðT 1  dðn  1ÞÞ < T 2 . Since 0  d(.)  dF therefore we have,
node on AV. Here we assumed Dw1 = 0 because AV node cannot
F
have effects on the SA node for refractoriness. For synchronization minðT 1  dðn  1ÞÞ ¼ T 1  d (15)
n
of two coupled pacemakers we must have

w < ’2 ðnÞ < 1 8n2N (8) maxðT 1  dðn  1ÞÞ ¼ T 1 (16)

n

Condition (8) is necessary for spontaneous synchronization Since we assumed T1 < T2 therefore we must have
because as mentioned before, the effect of SA node on AV node
should be excitatory which results in condition (8) to be necessary. F
wT 2 < T 1  d (17)
On the other hand, if condition (8) holds, it means that after each
discharge of SA node, there is an AV node discharge, which denotes To summarize, conditions (13) and (17) are sufficient for
synchronization. Therefore, condition (8) is necessary and synchronization. These conditions are all time independent and
sufficient for synchronization. Eq. (8) is time dependent and can be checked off-line.
should be checked for all times to identify whether two The spontaneous synchronization is investigated in Example 1
pacemakers are synchronized or not. To overcome this restriction, and the simulation results are shown in Fig. 7(a–c).
by relaxing the necessary condition, we obtain some sufficient Conditions (13) and (17) are obtained based on the assumption
conditions, which are time independent. that the SA node pulses reach the AV node immediately. In a real
Suppose d(n) = t2(n)  t1(n), then the Eq. (5) is rearranged as scenario, however, a time-delay must be taken into account.
follows: Assume that the SA node pulses reach the AV node after a constant
time-delay of td seconds, this changes Eq. (6) to:
t 1 ðnÞ ¼ t 1 ðn  1Þ þ T 1
(9) t 1 ðnÞ þ t d  t 2 ðn  1Þ
a ’2 ¼ (18)
t 2 ðnÞ ¼ t 2 ðn  1Þ þ T 2 þ ðT 1  dðn  1Þ  T 2 Þ T2
1w
 and Eqs. (9) and (10) change to:
a   a 
dðnÞ ¼ 1  dðn  1Þ þ 1  DT
1w 1w (10) t 1 ðnÞ ¼ t 1 ðn  1Þ þ T 1
a (19)
DT ¼ T 2  T 1 t 2 ðnÞ ¼ t 2 ðn  1Þ þ T 2 þ ðT 1 þ t d  dðn  1Þ  T 2 Þ
1w

Eq. (10) shows a delay dynamical equation (DDE) that is the delay  a   a  a
dðnÞ ¼ 1  dðn  1Þ þ 1  DT þ t (20)
between the AV node discharge times with respect to the SA node. 1w 1w 1w d
60 H. Gholizade-Narm et al. / Biomedical Signal Processing and Control 4 (2009) 57–66

and the condition (17) changes to:

F
wT 2 < T 1  d þ t d ; T 1 þ t d < T 2 (21)

To maintain simplicity, the derivations in the rest of this paper

are formulated for a hypothetical delay-less case. However,
impacts of such delay can easily be integrated into the equations Fig. 5. Discharge time of SA and AV pacemakers and external stimulus u(n).
by substituting ð1  ða=1  wÞÞDT by ð1  ða=1  wÞÞDT þ ða=1 
wÞt d and the (17) by (21). Since 0 < d(.) must hold, therefore according to (22) and (24) we
In the next section, the goal is how we can synchronize the two must have
nodes by applying a stimulus to the SA and AV nodes if the above a
conditions do not hold. 0< <1
1w
 (25)
aAV a 
5. Proposed method of synchronization ðT 2  STAVÞ < 1  DT
1  wAV 1w

According to aforementioned discussion, some pulses must be The new synchronization conditions are (17) and (25) except dF
applied to pacemakers in certain instances if they are not must be calculated from (24).
spontaneously synchronizable. At first, the sufficient conditions Using Eqs. (17), (24) and (25) and constraint (23), the boundary
for synchronization are obtained by assuming that we have access of stimulus pulse application interval can be determined. Example
only to the AV node. Then the above procedure will be repeated 2 discuses this part and the results of simulation show the
assuming we have access to the SA node. Finally, by assuming performance of proposed method in Fig. 8(a–c).
access to both nodes, the sufficient conditions of synchronization If one stimulus pulse cannot synchronize two pacemakers,
will be obtained. more pulses in the acceleration region of the AV node must be
It must be noted that the PRCs of external pulses are generally applied. Assume k stimulus pulses are applied to the AV node in the
different from the interaction PRC as it is considered in our acceleration region. The dynamical equations change as follows:
a aAV
ðT 1  dðn  1Þ  T 2 Þ þ
t 2 ðnÞ ¼ t 2 ðn  1Þ þ T 2 þ ðSTAV 1 þ STAV 2 þ . . . STAV k  kT 2 Þ
1w 1  wAV
 (26)
a   a  aAV
dðnÞ ¼ 1  dðn  1Þ þ 1  DT þ ðSTAV 1 þ STAV 2 þ . . . STAV k  kT 2 Þ
1w 1w 1  wAV

analysis. This implies that the magnitude and duration of external where STAVi, i = 1,2,,. . .,k are time intervals from stimulus pulses to
pulses can be considered as further PRC controlling parameters. t2(n) as given by
We assume that such parameters are known and given in this
study. wAV T 2 < STAV 1 < STAV 2 <    < STAV k < T 2 (27)

According to new equations, the fixed point dF is calculated as

5.1. Synchronization by having access to the AV node
ð1  ða=1  wÞÞDT þ ðaAV =1  wAV Þ
The cardiac blocking arrhythmias are caused by asynchrony
F ðSTAV 1 þ STAV 2 þ . . . STAV k  kT 2 Þ
between SA and AV nodes. One type of these arrhythmias happens d ¼ (28)
a=1  w
because of increasing the d (Blocking degree 2 type 1 or Mobitz 1 or
Wenkebach). In this type of arrhythmia, d increases until one of The new synchronization conditions are similar to the previous
impulses originated from SA node cannot reach ventricle. conditions (given by (17) and (25)) but (28) is used for evaluating dF
Asynchrony is caused by two major reasons: first, for the long P
and STAV is replaced by ki¼1 STAV i in (25). Using (28) and taking
distance of two intrinsic cycle length of two pacemakers, second, into consideration the synchronization conditions, the number of
for small parameters of PRC in which demonstrate the weak effects necessary stimulus pulses and their boundary of application
of SA node on AV node. To overcome the blocking arrhythmia, one intervals can be determined. An algorithm to determine the
way is by applying some stimulus pulses in proper instances to the number of stimulus pulses and their application intervals can be
AV node which increases its rate and therefore decrease d. suggested as follows:
Assuming one external pulse can synchronize two pacemakers
(Fig. 5), then the Eqs. (9) and (10) change as follows: Algorithm 1. Determining the number of pulses required and their
application intervals.
a aAV
ðT 1  dðn1Þ  T 2 Þ þ
t 2 ðnÞ ¼ t 2 ðn1ÞþT 2 þ ðSTAV T 2 Þ
1w 1  wAV
1. Start.
a a aAV 2. Check spontaneous synchronization conditions. If they are valid,
dðnÞ ¼ ð1  Þdðn  1Þ þ ð1  ÞDT þ ðSTAV  T 2 Þ
1w 1w 1  wAV then go to 7.
(22) 3. Consider wAV T 2 þ e; 0 < e < T 2  wAV T 2 for the first stimulus
pulse operating time.
where STAV is the time distance from external stimulus u(n) to
4. Apply stimuli pulses in their boundary interval.
t2(n) given by
5. If all conditions are satisfied, go to 7.
STAV ¼ uðnÞ  t 2 ðnÞ; wAV T 2 < STAV < T 2 (23) 6. Take the operating time of next stimulus pulse after minimum
time of former pulse and go to 4. (The minimum time is defined
F
According to (22), the fixed point d will be equal to by physiologist.)1
7. End.
ð1  ða=1  wÞÞDT þ ðaAV =1  wAV ÞðSTAV  T 2 Þ
dF ¼ (24)
a=1  w 1
Note that the excitation time should not exceed T2.
 H. Gholizade-Narm et al. / Biomedical Signal Processing and Control 4 (2009) 57–66 61

A sufficient condition for (33) is wT 2 < minn ðT 1  dðn  1Þ þ

ðdSA =wSA ÞSTSAÞ and maxn ðT 1  dðn  1Þ þ ðdSA =wSA ÞSTSAÞ < T 2 .
Since 0  d(.)  dF therefore we have
 
dSA F dSA
min T 1  dðn  1Þ þ STSA ¼ T 1  d þ STSA (34)
n wSA wSA
 
dSA dSA
max T 1  dðn  1Þ þ STSA ¼ T 1 þ STSA (35)
n wSA wSA
Fig. 6. Application of synchronizing stimulus pulse to SA node.
Therefore, we must have

F dSA
Remark: From the theoretical point of view, smaller e has wT 2 < T 1  d þ STSA
wSA (36)
greater effect and fewer number of necessary external pulses for dSA
T1 þ STSA < T 2
synchronization will be required. But from the practical point of wSA
view, smaller e may cause wrong results because a little
The synchronization conditions are (32), (36), and (31) is used
uncertainty in PRC or little change in T1 and T2 (which is
for evaluating dF. This part is exemplified in Example 4 and
unavoidable for automatic nervous systems influences) may move
simulation results are illustrated in Fig. 10(a and b).
the pulse to the delay region instead of applying in the acceleration
If one stimulus pulse cannot synchronize two pacemakers,
one. Therefore, from a practical point of view, it is better to choose
more pulses in delay region of the SA node must be applied.
e ¼ ðT 2  wAV T 2 =2Þ.
Assume k stimulus pulses are applied to the SA node in the delay
The usage of Algorithm 1 is demonstrated in Example 3 and
region. We have
Fig. 9(a–c).
dSA
t 1 ðnÞ ¼ t 1 ðn  1Þ þ T 1 þ ðSTSA1 þ STSA2 þ . . . þ STSAk Þ
5.2. Synchronization by having access to the SA node wSA
a dSA
t 2 ðnÞ ¼ t 2 ðn  1Þ þ T 2 þ T1 þ ðSTSA1 þ STSA2
The SA node plays a crucial and pivotal role in cardiac 1w wSA
!
synchronization process and it is better to coordinate AV node
with it. But, when there is no access to the AV node, the þ . . . þ STSAk Þ  dðn  1Þ  T 2
synchronization pulses should be applied to the SA node to avoid (37)
 a   a  d
blocking. These pulses cause delay in the SA discharge time and dðnÞ ¼ 1  dðn  1Þ þ 1  DT  SA
allow the AV node to coordinate itself with the SA node. In other 1w 1w wSA
words, synchronizing stimulus pulses must be applied in the delay a d
ðSTSA1 þ STSA2 þ . . . þ STSAk Þ þ  SA
region of the SA node (Fig. 6). 1  w wSA
The dynamical equations will be changed, if one stimulus pulse
applied in the delay region of the SA node, as follows: ðSTSA1 þ STSA2 þ . . . þ STSAk Þ

where
dSA
t 1 ðnÞ ¼ t 1 ðn  1Þ þ T 1 þ STSA
wSA 0 < STSAk < STSAk1 <    < STSA1 < wSA T 1
 
a d Minimum Necessary Time < STSAi  STSAiþ1 ; i ¼ 1; 2;    ; k  1
t 2 ðnÞ ¼ t 2 ðn  1Þ þ T 2 þ T 1 þ SA STSA  dðn  1Þ  T 2
1w wSA (38)
 a   a 
dðnÞ ¼ 1  dðn  1Þ þ 1  DT The reason for the constraint (38) is that two pacemakers are
1w  1  w
dSA a dSA desired to become synchronized with the least number of stimulus
 STSA þ STSA pulses. The fixed point is obtained as
wSA 1  w wSA
ð1  ða=1  wÞÞDT  ðdSA =wSA ÞðSTSA1 þ . . . þ STSAk Þ
where STSA is the constant time distance of stimulus pulse from
t1(n). þ ða=1  wÞ  ðdSA =wSA ÞðSTSA1 þ . . . þ STSAk Þ
dF ¼ (39)
a=1  w
STSA ¼ uðnÞ  t 1 ðnÞ; 0 < STSA < wSA T 1 (30)
The new synchronization conditions are similar to previous
The fixed point is now obtained as conditions given by (32) and (36) except that (39) is used for
P
evaluation of dF and STSA is replaced by ki¼1 STSAi in (32) and (36).
ð1  ða=1  wÞÞDT  ðdSA =wSA ÞSTSA Using (39) and considering aforementioned conditions, the
þ ða=1  wÞððdSA =wSA ÞSTSAÞ number of stimulus pulses and their boundary of application
dF ¼ (31) intervals can be determined. The algorithm, which determines the
a=1  w
number of the stimulus pulses and application interval, is proposed
Since 0 < d(.) must hold, therefore according to (29) and (31) we as follows:
must have
Algorithm 2. Determining the number of pulses required and their
a application intervals.
0< <1
 1  w  
a  d a dSA (32)
1 DT > SA STSA  STSA 1. Start.
1w wSA 1  w wSA
2. Check conditions for spontaneous synchronization. If they are
On the other hand, to satisfy (8) we must have satisfied, go to 7.
3. Consider STSA ¼ wSA T 1  e; 0 < e < wSA T 1 for the first stimulus
T 1  dðn  1Þ þ ðdSA =wSA ÞSTSA pulse application time.
w< <1 (33)
T2 4. Apply stimulus pulse in its boundary interval.
62 H. Gholizade-Narm et al. / Biomedical Signal Processing and Control 4 (2009) 57–66

5. If all conditions are satisfied, go to 7. The synchronization conditions are now (44) and (46) and (43)
6. Take the application time of next stimulus pulse after minimum is used for evaluating dF. By using (43) and the aforementioned
time of former pulse and go to 4. (The minimum time is defined conditions (44) and (46) and the constraint (42), the boundaries of
by physiologist.)2 stimuli pulses are achieved. Depending on the intrinsic cycle
7. End. length and parameters of PRC, it may not need any stimulus pulse.
They may be synchronized by applying only one stimulus pulse or
more, or they may not be synchronizable. In the following, we are
Remark: Again e near zero is preferred from theoretical point of
going to obtain the number and the operating intervals of stimuli
view but if there is a little uncertainty in PRC parameters or little
pulses. Suppose k1 delay stimulus pulses are applied to the SA node
change in T1 and T2, it is possible that external pulses fall in the
and k2 accelerate to the AV node. Therefore, the dynamical
wrong region. Therefore, from a practical point of view, it is better
equations change to:
to choose e equal to wSA T 1 =2. Algorithm 2 is used in Example 5 and
the results are shown in Fig. 11(a and b). dSA
t 1 ðnÞ ¼ t 1 ðn  1Þ þ T 1 þ ðSTSA1 þ STSA2 þ . . . þ STSAk1 Þ
wSA
5.3. Synchronization by having access to both SA and AV nodes a  d
t 2 ðnÞ ¼ t 2 ðn  1Þ þ T 2 þ T 1 þ SA ðSTSA1 þ STSA2
1w wSA
In the case of having access to both nodes, by applying delay  aAV
þ . . . þ STSAk1 Þ  dðn  1Þ  T 2 þ ðSTAV 1 þ STAV 2
stimulus pulses to the SA node and accelerate the pulses to the AV 1  wAV
node, one can synchronize two pacemakers under more restricted þ . . . STAV k2  k2 T 2 Þ
conditions, whereby they could not be synchronized by having  a   a  d
only access to one pacemaker. The goal of this section is to obtain dðnÞ ¼ 1  dðn  1Þ þ 1  DT  SA ðSTSA1
1w 1w wSA
these conditions. Suppose one delay stimulus pulse is applied to
the SA node and an accelerated pulse to the AV node, to aAV
þ STSA2 þ . . . þ STSAk1 Þ þ ðSTAV 1 þ STAV 2
synchronize the two pacemakers. 1  wAV
Dynamical equations of SA and AV nodes discharge time and a dSA
þ . . . STAV k2 k2 T 2 Þþ  ðSTSA1 þSTSA2 þ. . . þ STSAk1 Þ
DDE is as follows: 1  w wSA
dSA where
t 1 ðnÞ ¼ t 1 ðn  1Þ þ T 1 þ STSA
wSA
aAV a 0 < STSAk1 < STSAk1 1 <    < STSA1 < wSA T 1
t 2 ðnÞ ¼ t 2 ðn  1Þ þ T 2 þ ðSTAV  T 2 Þ þ (40) Minimum Necessary Time < STSAi  STSAiþ1 ; i ¼ 1; 2;    ; k1  1
1  wAV 1w
  wAV T 2 < STAV 1 < STAV 2 <    < STAV k2 < T 2
dSA
T1 þ STSA  dðn  1Þ  T 2 Minimum Necessary Time < STAV iþ1  STAV i ; i ¼ 1; 2;    ; k2  1
wSA
(48)
 a   a  d
dðnÞ ¼ 1  dðn  1Þ þ 1  DT  SA STSA Thus, the constraint (48) is added to the conditions (44) and
Pk1
1w 1w wSA
  (46) except that STSA and STAV must be replaced by i¼1 STSAi and
aAV a dSA Pk2
þ ðSTAV  T 2 Þ þ STSA (41) i¼1 STAV i respectively in (44) and (46). According to these
1  wAV 1  w wSA conditions and the constraint (48), one can determine the number
where of necessary stimuli pulses and the intervals in which they should
be applied. This part has been discussed in Example 6 and Fig. 12(a
STSA ¼ uSA ðnÞ  t 1 ðnÞ; 0 < STSA < wSA T 1 and b).
(42)
STAV ¼ uAV ðnÞ  t 2 ðnÞ; wAV T 2 < STAV < T 2
6. Simulation studies
The fixed point dF obtained from DDE is

B In this section, we illustrate our proposed way with several

dF ¼  numerical examples.
A
 a  aAV
B¼ 1 DT þ ðSTAV  T 2 Þ (43) Example 1. Suppose for SA node, we have T1 = 1 s and PRCSA = [d-
1w 1  wAV
  SA,aSA,wSA] = [0.4,0.4,0.5] and for AV node T2 = 1.7 s and PRCAV = [-
dSA a dSA
 STSA þ STSA dAV,aAV,wAV] = [0.26,0.26,0.5] and the interaction PRC as
wSA 1  w wSA
PRC = [d,a,w] = [0.42,0.42,0.5]. Checking the conditions for sponta-
For d(.) > 0 according to (41) and (43) we must have neous synchronization we obtain:
a a F F
0< <1 0< ¼ 0:84 < 1; d ¼ 0:1333 ) wT 2 ¼ 0:85 < T 1  d
1w (44) 1w
B>0 ¼ 0:8667:
On the other hand, to satisfy (8), we must have
Thus, sufficient conditions hold and it is expected that the two
T 1  dðn  1Þ þ ðdSA =wSA ÞSTSA pacemakers become spontaneously synchronized. Simulation
w< <1 (45)
T2 results are shown in Fig. 7(a–c) in which it can be seen that after
A sufficient condition for (45) is similar to (33) as follows: a delay from SA depolarization, AV node depolarizes as well.

F dSA Example 2. Suppose a = 0.2, d = 0.3 and other parameters are the
wT 2 < T 1  d þ STSA
wSA same as previous example. In this case, dF = 1.05 and we get
(46) F
dSA wT 2 ¼ 0:85 > T 1  d ¼ 0:05. This indicated that condition (17)
T1 þ STSA < T 2
wSA has not been satisfied. Moreover, we have ’F2 < 0 which is outside
the acceptable range and therefore, the two pacemakers cannot be
2
Note that the stimulus time should not exceed wSA T 1 and not less than zero. spontaneously synchronized without an external stimulus pulse.
 H. Gholizade-Narm et al. / Biomedical Signal Processing and Control 4 (2009) 57–66 63

Fig. 7. Spontaneous synchronization (a) Convergence to SA cycle length of AV node (b) Discharge time of two pacemakers (c) confirmation of conditions (8).

Fig. 8(a) illustrates that after two times of SA depolarization, AV
can depolarize only one time, which indicates a 2:1 blocking. As it
can be seen, the two pacemakers cannot be synchronized without
an external stimulus pulse. Here we apply a stimulus pulse at a
proper time to synchronize the two pacemakers. According to
Eqs. (17) and (25) and constraint (23) on STAV, we have
0.89231 < STAV < 1.0077 (the maximum value of STAV is restricted
by (17) and its minimum by (25)), suppose STAV = 1 s then
dF = 0.14 s. Simulation results are shown in Fig. 8(b and c) in which
part (b) shows convergence of the AV node to SA node cycle length
and the discharge time is shown in part (c).
Example 3. Changing d and a in Example 2 to d = 0.25 and a = 0.15,
and checking condition (17) reveals that since dF = 0.16 for
F
SAV = 0.85 and wT 2 ¼ 0:85 > T 1  d ¼ 0:84, condition (17) still is
not satisfied. Next, following the same approach as Example 2 and
applying a single pulse, Fig. 9(a) shows that a single pulse unlike
previous example is not capable of synchronizing the nodes. This is
because condition (17) is sufficient and not necessary; therefore it
is possible that the two pacemakers are not synchronized by
applying one stimulus. It is however expected that by applying
more stimulus pulses, the two pacemakers be synchronized. Using
Algorithm 1 we can find the number of necessary pulses and

Fig. 8. 2:1 Blocking (SA:AV) due to lack of synchronization between the two pacemakers and synchronization of the two pacemakers by applying a stimulus pulse to AV node
(a) 2:1 blocking, (b) convergence of two pacemakers, (c) SA and AV nodes discharge times.
64 H. Gholizade-Narm et al. / Biomedical Signal Processing and Control 4 (2009) 57–66

Fig. 9. Lack of synchronization between the two pacemakers and their synchronization after applying two stimulus pulses to AV node (a) lack of synchronization due to
applying inadequate number of pulses, (b) convergence of two pacemakers, (c) discharge time of SA and AV nodes.

proper times to apply them. For this problem, Algorithm 1 gives
STAV1 = 1, STAV2 = 1.52, and dF = 0.108. Fig. 9(b and c) show that the
two pacemakers have been synchronized using two pulses applied
to the AV node.
Example 4. Consider Example 2 in which the oscillators became
synchronized by applying a stimulus pulse to the AV node. In this
example, we would like to achieve synchronization by applying a
stimulus pulse to the SA node. The boundary of applying stimulus
pulse can be obtained according to the pacemakers’ parameters,
conditions (32) and (36) and constraint (30). Let us assume
STSA = 0.47 for simulation, which results dF = 0.486 and the afore-
mentioned conditions hold. Fig. 10 summarizes the simulation
results for this case and confirms that the synchronization is
achieved.
Example 5. Consider the two pacemakers in Example 3. As in
Example 4, we would like to achieve synchronization by applying
a stimulus pulse to the SA node. The two pacemakers cannot
be synchronized even when STSA = 0.49, which provides the
F
maximum effect, since dF = 0.7 and wT 2 ¼ 0:85 > T 1  d þ
ðdSA =wSA ÞSTSA ¼ 0:7. Therefore condition (36) is not satisfied.
For this case since ’F2 < 0:37, which is less than 0.5, suggests that
more pulses is required for synchronization. Using Algorithm 2 we
find that we need two pulses for synchronization with STSA1 = 0.4
and STSA2 = 0.3 which gives ’F2 ¼ 0:726, dF = 0.3267. Fig. 11 con-
firms that the desired synchronization was achieved.
Example 6. Consider two pacemakers with parameters
PRC = [d,a,w] = [0.2,0.1,0.5], which describes the mutual interac-
tions between the nodes, T1 = 1 and PRCSA = [dSA,a-
SA,wSA] = [0.2,0.25,0.5] to define the SA node and T2 = 1.9 and
PRCAV = [dAV,aAV,wAV] = [0.2,0.2,0.5] to identify the AV node. As
we can see from these sets of PRCs parameters, the magnitude
of d and a (delay and acceleration slops) are very small. These
express little mutual interaction and little external pulse effects.
Thus it makes the synchronization from one node harder or even
impossible. In this case, one can apply some stimulus pulses to the
SA and some to the AV nodes together to synchronize them.
According to sufficient conditions and constraints for synchroniza-
tion, we can apply a set of two pulses to the AV and another two
pulses to the SA node as described by the following set of para-

Fig. 10. Synchronization of two pacemakers by applying a stimulus pulse to SA node (a) convergence of cycle length of the two pacemakers, (b) SA and AV nodes discharge
time.
 H. Gholizade-Narm et al. / Biomedical Signal Processing and Control 4 (2009) 57–66
Fig. 11. Synchronization of two pacemakers by applying two stimulus pulses to SA node (a) convergence of cycle length of two pacemakers (b) SA and AV nodes discharge
times.
Fig. 12. Synchronization by applying 2 stimulus pulses to SA node and 2 stimulus pulses to AV node (a) convergence of beat cycle of two pacemakers on 1.2 s (b) SA and AV
nodes discharge times.
meters: STSA1 = 0.3, STSA2 = 0.2 and STAV1 = 1, STAV2 = 1.5. This
would give us: dF = 0.2, and ’F2 ¼ 0:526. Simulation results confirm
that synchronization has been achieved as seen by Fig. 12.
7. Conclusion
Synchronization analysis of cardiac pacemaker cells is often
performed in two ways. The first method uses nonlinear oscillators
such as Van der Pol, FitzHugh–Nagumo, or Hodgkin–Huxley like
models and the second method is based on the PRC. The major
drawback of the first method is the high magnitude of the
synchronization signal. In addition, this method needs all state
variables. Using the second method based on the PRC has
advantages such as (1) no requirement to solving nonlinear
differential equations, (2) only the stroke time of action potential is
sufficient to be sampled, (3) synchronization signals are stimulus
pulses which have small amplitudes and (4) the stimulus pulse
generation is much simpler than generation of the signal
determined by nonlinear tools. The major disadvantage of the
second method is the need to have the PRC parameters. Such
parameters play an important role in synchronization and are
obtained using stimulations with short stimuli of an oscillator.
In this paper, we proposed novel time-independent conditions
to check whether two pacemakers synchronize spontaneously or
not. These conditions are results of stability analysis of proposed
dynamical equations and PRC features. Furthermore, they did not
synchronize spontaneously, how many external pulses were
needed and when must they be applied to pacemakers to
synchronize them? This research establishes sufficient conditions
for spontaneous synchronization and the use of external stimulus
pulses to achieve synchronization. If these conditions are satisfied
by the two pacemakers, their synchronization is guaranteed. The
number of necessary pulses and proper instances for application of
65
pulses were also provided. The presented examples illustrate
application of the theoretical results.
References
[1] Ikeda Noriaki, Model of bidirectional interaction myocardial pacemakers based on
the phase response curve, Biol. Cybern. 43 (1982) 157–167.
[2] I. Harary, B. Farley, In vitro studies on single beating rat heart cells: II. Intracellular
communication, Exp. Cell. Res. 29 (1963) 466–474.
[3] G.E. Mark, F.F. Strasser, Pacemaker activity and mitosis in cultures of newborn rat
heart ventricle cells, Exp. Cell. Res. 44 (1966) 217–233.
[4] K. Goshima, Synchronized beating of and electronic transmission between myo-
cardial cells mediated by heterotypic strain cells in monolayer culture, Exp. Cell.
Res. 58 (1969) 420–426.
[5] R.L. Dehaan, R. Hirakow, Synchronization of pulsation rates in isolated cardiac
myocytes, Exp. Cell. Res. 70 (1972) 214–220.
[6] T. Sano, T. Sawanobori, H. Adaniya, Mechanism of rhythm determination among
pacemaker cells of the mamalian sinus node, Am. J. Physiol. 235 (1978) H379–
H384.
[7] C. Antzelovich, J. Jalife, G.K. Moe, Electronic modulation of pacemaker activity.
Further biological and mathematical observations on the behavior of modulated
parasystole, Circulation 66 (1982) 1225–1232.
[8] J. Jalife, K. Moe, Effect of elctronic potentials on pacemaker activity of canine
purkinje fibers in relation to parasystole, Circ. Res. 39 (1976) 801–808.
[9] V.S. Reiner, C. Antzelovic, Phase resetting and annihilation in a mathematical
model of sinus node, Am. J. Physiol. 249 (1985) H1143–H1153.
[10] M.R. Guevara, A. Shrier, L. Glass, Phase-locked rhythms in periodically stimulated
heart cell aggregates, Am. J. Physiol. 254 (1988) H1–H10.
[11] J. Jalife, A.J. Hamilton, V.R. Lamanna, G.K. Moe, Effects of current flow on pace-
maker activity of the isolated kitten sinoatrial node, Am. J. Physiol. 238 (1980)
H307–H316.
[12] D.M. Wei, G. Yamada, T. Musha, H. Tsunakawa, T. Tsutsumi, K. Harumi, Computer
simulation of supraventricular tachycardia with the Wolff–Parkinson–White
syndrome using three-dimensional heart models, J. Electrocardiol. 23 (1990)
261–273.
[13] O. Okazaki, D. Wei, K. Harumi, Three dimensional excitation sequence during
Torsade de point in the cardiac computer model with M cells. 2nd International
Conference on Bioelectromagnetism, February 1988, 131–132.
[14] A.-S. Sarit, A. Solang, A phase response curve based model: effect of vagal and
sympathetic stimulation and interaction on a pacemaker cell, J. Theor. Biol. 192
(1998) 567–579.
66 H. Gholizade-Narm et al. / Biomedical Signal Processing and Control 4 (2009) 57–66
[15] A.-S. Sarit, A. Solang, Simulation of atrial activity by a phase response curve based
model of a two-dimensional pacemaker cells array: the transition from a normal
activation pattern to atrial fibrillation, Biol. Cybern. 80 (1999) 141–153.
[16] A.-S. Sarit, A. Solang, A simulation of the SA node by phase response curve-based
model of a two-dimensional pacemaker cells array, IEEE Trans. Biomed. Eng. 47
(4) (2000).
[17] D.G. Tsalikakis, H.G. Zhang, D.I. Fotiadis, G.P. Kremmydas, Ł.K. Michalis, Phase
response characteristics of sinoatrial node cells, Comput. Biol. Med. 37 (2007) 8–
20.
[18] H. Zhang, A.V. Holden, I. Kodama, H. Honjo, M. Lei, T. Varghese, M.R. Boyett,
Mathematical models of action potentials in the periphery and center of
the rabbit sinoatrial node, Am. J. Physiol. Heart Circ. Physiol. 279 (2000)
H397–H588.
[19] J. Jalife, C. Anzelevitch, Phase resetting and annihilation of pacemaker activity in
cardiac tissue, Science 206 (1979) 695–697.
[20] L. Glass, M.C. Mackey, From Clock to Chaos: The Rhythms of life, Princeton Univ.
Press, New Jersey, 1988, pp. 98–118.