164
IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 53, NO. 2, FEBRUARY 2006
A Comparison of 1-D Models of Cardiac Pacemaker Heterogeneity Shaun L. Cloherty*, Member, IEEE, Socrates Dokos, and Nigel H. Lovell, Senior Member, IEEE
Abstract—In this paper, we investigate the role of sinoatrial node (SAN) cellular heterogeneity in two key aspects of normal cardiac pacemaker function: frequency entrainment of the SAN, and propagation of excitation into the atrial tissue. Using detailed ionic models of electrical activity in SAN and atrial myocytes, we have formulated a number of one-dimensional models of SAN heterogeneity based on discrete-region (in which central and peripheral SAN type cell are separated into discrete regions), gradient and mosaic models of SAN organization. Each of the different models were assessed on their ability to achieve frequency entrainment of the SAN and activation of the adjoining atrial tissue in the presence of both uniform and linearly increasing conductivity profiles. Simulation results suggest that the gradient model of SAN heterogeneity, in which cells display a smooth variation in membrane properties from the center to the periphery of the SAN, produces action potential waveshapes and a site of earliest activation consistent with experimental observations in the intact SAN. The gradient model also achieves frequency entrainment of the SAN more easily than other models of SAN heterogeneity. Based on these results, we conclude that the gradient model of SAN heterogeneity, in the presence of a uniform conductivity profile, is the most likely model of SAN organization. Index Terms—Action potential heterogeneity, frequency entrainment, mathematical modeling, monodomain model, sinoatrial node.
I. INTRODUCTION
U
NDER normal conditions, electrical activation of the mammalian heart is initiated in a small region of specialized pacemaker cells known as the sinoatrial node (SAN). Regional variation in cellular electrical properties (both ionic and structural) is a widely accepted characteristic of the intact SAN, which has been well studied in the rabbit [1]. However, little attention has been given to its contribution to both normal and abnormal pacemaker function. In a recent simulation study, we reported that heterogeneity in action potential (AP) waveshape assisted frequency entrainment of electrically coupled pacemaker cell pairs [2]. It was proposed that regional variation in AP characteristics within the intact SAN provides an important mechanism underlying pacemaker synchronization [2]. In this paper we formulate a number of computational models of heterogeneity in the rabbit SAN and investigate the relative merits of each model in facilitating the two main
Manuscript received November 29, 2004; revised May 22, 2005. Asterisk indicates corresponding author. *S. L. Cloherty is with the Graduate School of Biomedical Engineering, University of New South Wales, Sydney, NSW 2052, Australia (e-mail:
[email protected]). S. Dokos and N. H. Lovell are with the Graduate School of Biomedical Engineering, University of New South Wales, Sydney, NSW 2052, Australia. Digital Object Identifier 10.1109/TBME.2005.862538
functions of the SAN: maintaining a robust synchronous rhythm, and electrical activation of the adjoining atrial tissue. Early studies of the SAN based on histological, ultrastructural and electrophysical observations describe a discrete-region model of SAN organization (for a review see [3]). According to this discrete-region model, the SAN is composed of a compact central region of primary pacemaker cells surrounded by a region of transitional cells forming a buffer zone between the primary pacemaker cells and the working myocardium. The cardiac AP is initiated in the compact central region of the SAN and propagates outwards through the peripheral region and into the atrial tissue. In a later study, Kodama and Boyett [1] observed a variation in AP characteristics in small isopotential balls of tissue in diameter) isolated from known spatial locations ( within the SAN. They concluded that the variation in AP waveshape was due to a genuine transition in ion channel expression, and proposed the gradient model of SAN heterogeneity. More recently, Verheijck et al. [4] identified three morphologically different types of SAN cells in addition to atrial myocytes in rabbit SAN tissue. After enzymatic isolation, these three cell types appeared to exhibit no significant variation in AP characteristics. They hypothesized that the observed variation in AP waveshape from the center to the periphery of the SAN was due to a gradual increase in atrial cell numbers toward the periphery. This interpretation forms the basis of the mosaic model of SAN heterogeneity. In this paper, we attempt to identify the most plausible model of SAN heterogeneity. We employ a biophysically detailed ionic model of a single SAN myocyte, able to exactly reproduce AP recordings from both the center and periphery of the rabbit SAN. With appropriate parameters, this model is also able to produce AP waveshapes characteristic of the intermediate region between the center and the periphery of the SAN. Several one-dimensional (1-D) models of the SAN are formulated, based on the three conceptual models of SAN heterogeneity described above, namely the discrete-region model, the gradient model and the mosaic model. We present results from a number of simulations aimed at assessing the relative merits of each conceptual model of SAN heterogeneity and attempt to identify a number of mechanisms which likely underly normal SAN pacemaker function. II. METHODS A. The Single Cell Models The membrane ionic models of SAN tissue were based on a generic ionic model of isolated rabbit SAN myocytes [5]. This
0018-9294/$20.00 © 2006 IEEE
CLOHERTY et al.: A COMPARISON OF 1-D MODELS OF CARDIAC PACEMAKER HETEROGENEITY
single cell model includes formulations for 12 membrane currents together with dynamic changes in ionic concentrations. The default parameters of the generic model were selected to faithfully reproduce published voltage clamp data applicable to the rabbit SAN. In order to reproduce the observed heterogeneity in AP waveshapes from rabbit SAN, a subset comprising 173 of the generic model parameters were then fine-tuned using a nonlinear optimization routine [5], [6], to fit AP recordings from central and peripheral regions. Throughout the optimization process, the model parameters were subject to tight constraints designed to ensure reasonable correlation with the default values and that the peak ion current amplitudes remain in close agreement with experimental observations. The resulting central and peripheral SAN cell models provided a basis for formulating a smooth transition in cell model parameters in the intermediate region. Model parameters which produced AP waveshape characteristics [namely: maximum diastolic potential (MDP), overshoot potential (OS), upstroke velocity (UV), action potential duration (APD), and cycle length (CL)] representative of the intermediate region were identified in a separate optimization procedure as described previously [5]. Atrial tissue was simulated using the Earm, Hilgemann, and Noble (EHN) model of a single rabbit atrial cell [7], [8]. B. One-Dimensional Model of the Intact Cardiac Pacemaker The SAN and adjoining atrial tissue was approximated as a highly idealized 1-D cable, consisting of a region of SAN tissue coupled to a region of atrial tissue as illustrated schematically in Fig. 1. Electrical activity in the SAN and atrial tissue was determined by the monodomain equation (1) ), dewhere denotes the tissue conductivity ( denotes the cell notes the transmembrane potential (mV), ), denotes the cell specific surface-to-volume ratio ( membrane capacitance ( ), denotes time (s), and denotes the sum of the transmembrane ionic currents ( ), the kinetics of which are governed by the underlying single cell ionic models. Neumann (zero-flux) boundary conditions were imposed at both ends of the 1-D model, i.e., (2) A conservation of flux condition given by (3) was imposed at the SAN-atrial border since the tissue conductivity ( ) was discontinuous at (3) The length of the SAN region ( ) was set to 1.5 mm, roughly in agreement with the distance from the center to the periphery of the SAN perpendicular to and in the direction of the crista terminalis (CT) in the rabbit [1], [9]. The length of the atrial tissue region ( ) was also set to 1.5 mm. While considerably
165
Fig. 1. The idealized 1-D model of the SAN and adjoining atrial tissue. The total length of the preparation was 3 mm, with l = l = 1:5 mm. Zero flux boundary conditions were imposed at both ends of the model.
shorter than the spatial scale of the right atrium in the rabbit, this length corresponds to roughly 1.5 space constants and was sufficient to approximate the electrotonic load imposed on the SAN by the atrial tissue without any adverse effect on the SAN from the zero flux boundary condition imposed at the end of the preparation. There is considerable experimental evidence to suggest that the specific membrane capacitance in cardiac tissue is largely independent of cell type and may reasonably be assumed to be ap[10]. Therefore, was assigned proximately 0.01 throughout both the SAN and a uniform value of 0.01 atrial regions of the 1-D models formulated in this paper. in Cells of the rabbit SAN are reported to be roughly 8 in length [9]. Atrial cells are reported diameter and 25–30 in diameter and 100 to be somewhat larger, approaching 20 in length [11]. Therefore, assuming that cells of the SAN and atria are roughly cylindrical results in estimated surface-toand 200 respectively. volume ratios of 500 C. Modeling Sinoatrial Node Heterogeneity Three models of SAN heterogeneity, namely the discrete-region, gradient and mosaic models, were formulated within the framework of the 1-D model of the SAN described above. The Discrete-Region Model: In the Discrete-Region model, the SAN region of the 1-D model was divided into two discrete and . Computational areas, nodes within these two regions were assigned central and peripheral SAN cell characteristics respectively. The size of the central ) was consistent with the data of Koregion ( dama et al. [1], who observed central like AP waveshapes in only in diameter) prepared from strands of 1 or 2 balls ( tissue running from the center to the periphery of the SAN. The Gradient Model: In the Gradient model, a smooth transition in AP waveshape characteristics was imposed in the SAN region of the 1-D model. Computational nodes in the SAN region were assigned SAN model parameters to produce a linear variation (with respect to ) in AP waveshape characteristics to those of a peripheral from those of a central SAN cell at SAN cell at . The Mosaic Model: In the mosaic model, nodes in the SAN region of the 1-D model were randomly designated as being of either SAN or atrial type. The probability of a node being designated of atrial type increased linearly with , from 0.0 at to 1.0 at . The average probability of a node being of atrial ) and type was, therefore, 0.16 in the central portion ( ) of the SAN re0.66 in the peripheral portion ( gion. These average probabilities compare reasonably well with the proportions of atrial type cells observed by Verheijck et al. , ) and in the in the dominant pacemaker region ( , ) of the rabbit [4]. CT (
166
IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 53, NO. 2, FEBRUARY 2006
The following four variations of the mosaic model were examined in this paper. 1) The Mosaic-CA model, in which nodes designated as being of SAN type were simulated using model parameters corresponding to the central type SAN cell. Those designated as being of atrial type were simulated using the single rabbit atrial cell model employed in the atrial region of the 1-D model. 2) The Mosaic-PA model, which was formulated in the same manner as above, substituting the peripheral SAN cell model in place of the central SAN cell model. 3) The Mosaic-CP model, which employed the central SAN cell model at nodes designated as being of SAN type, and the peripheral SAN cell model at nodes designated as being of atrial type. 4) The Mosaic-CPA model, which employed all three cell types. Nodes designated as being of SAN type were randomly assigned model parameters corresponding to either central or peripheral SAN type cells. Nodes designated as being of atrial type were simulated using the atrial cell model. In the Mosaic-CP variation of the mosaic model, nodes designated as being of atrial type and assigned parameters corresponding to the peripheral SAN cell model may be thought of as representing small regions of tissue containing both SAN and atrial cells with a predominance of atrial type cells. Similarly, nodes designated as being of SAN type and assigned parameters corresponding to the central SAN cell model may be thought of as representing small regions of tissue containing a predominance of SAN type cells. In the Mosaic-CPA model, the probability of a node being ) decreased monotonidesignated as central SAN type ( to 0.0 at according to cally with , from 1.0 at (4) where i.e.,
is the normalized distance from the center of the SAN, , and determines the spatial rate of decay of with distance from the center of the SAN. The parameter was set to 6.3884 such that the overall probability of a node being designated as central SAN type was reduced to at . Fig. 2 shows the probability distribution for each cell type in the SAN region of the Mosaic-CPA model as a function of distance from the center of the SAN. D. SAN–Atrial Conductivity Profiles Whilethereislittleexperimental informationregarding thedistribution of gap junctions within the SAN, histological studies suggest the density of gap junctions in atrial tissue may be as much as 10 times higher than in the SAN [12]. Although the exact mechanism remains unclear, there exists some evidence to suggest a spatial gradient in conductivity from the center to the periphery of the SAN [11]. The 1-D models formulated in this paper, therefore, employ two idealized conductivity profiles throughout the SAN region. The first imposes a uniform conductivity throughout the SAN (Fig. 3, top panel), whilst the second establishes a linear increase in conductivity from the center to the periphery of the SAN (Fig. 3, bottom panel). . In simThe atrial conductivity ( ) was set to 250 ulations of a 1-D strand of atrial tissue, this conductivity yielded
Fig. 2. Probability distributions for central SAN, peripheral SAN and atrial type cells in the SAN region of the Mosaic-CPA model.
a space constant of approximately 1.0 mm and an intrinsic con. Removal duction velocity (CV) of approximately 0.5 of the atrial tissue from the intact SAN is known to cause a shift in the site of earliest activation away from the center and toward the periphery of the SAN [13]. In order to assess the ability of each model to reproduce this shift and also to determine the effect of the atrial load on entrainment of the SAN, a number of simulations were performed in which the SAN region was isolated from the atrial tissue region by setting the atrial conductivity ( ) to zero. In this configuration, the conservation of flux condition [see (3)] reduced to the familiar Neumann or zero-flux boundary condition imposed at the ends of the 1-D model. E. Frequency Entrainment of the Sinoatrial Node In each model of SAN heterogeneity described above, cells within the SAN region exhibit different intrinsic CLs. At low values of SAN conductivity these cells remain essentially independent. At sufficiently large SAN conductivities, cells achieve frequency entrainment at a common CL. A number of simulations were performed to determine the minimum conductivity required for frequency entrainment of each model of SAN heterogeneity described above. The criterion for assessing the entrainment state of the SAN is described below. The critical con, was determined for each model ductivity, denoted as with the uniform conductivity profile both in the absence and in the presence of the atrial tissue load. The process was repeated using the linear conductivity profile ) set to with the conductivity at the center of the SAN region ( . This value was approximately 25% of the min0.01 imum critical conductivity for entrainment of the SAN with the uniform conductivity profile (see Section III-B). In the case of the linear conductivity profile, the critical conductivity for entrainment was defined as the minimum conductivity required at the pe) and was denoted as . riphery of the SAN region ( In each configuration, the model was simulated for 5 s to stabilise any initial transients. The model was then simulated for a further 10 s during which the entrainment state of the SAN region was assessed. With the uniform conductivity profile, both the gradient and mosaic models took considerably longer to approach steady state. In these configurations, the model was simulated for an initial period of 20 s before commencing the 10 s observation period.
CLOHERTY et al.: A COMPARISON OF 1-D MODELS OF CARDIAC PACEMAKER HETEROGENEITY
167
Fig. 3. The uniform (top) and linear (bottom) SAN–Atrial conductivity profiles employed in the 1-D models of the SAN and adjoining atrial tissue. A conservation of flux condition was imposed at the boundary between the SAN and atrial tissue regions where the conductivity is discontinuous.
For the purposes of this paper, a given configuration was deemed to be entrained if, for each cycle in the 10 s observation period, the range of CLs observed within the SAN region was less than 5% of the mean CL in each cycle. The CL at a given node was defined to be the interval between moments of maximal UV. The and ) were then compared critical conductivities ( in order to assess the relative ease with which each of the models of SAN heterogeneity achieved frequency entrainment. F. Activation of the Adjoining Atrial Tissue A series of simulations were performed for each model of SAN heterogeneity to determine the minimum SAN conductivity required for propagation of the AP from the SAN into the adjoining atrial tissue. The criterion for determining successful activation of the atrial tissue by the SAN is described below. Simulations were performed first with the uniform conductivity profile described above. The minimum SAN conductivity required for activation of the atrial tissue, referred to as the critical conductivity for activation, was denoted as . The procedure was then repeated with the linear conductivity profile, with the conductivity at the center of the SAN region ( ) set to 0.01 . As for the entrainment simulations described above, the critical conductivity for activation with the linear conductivity profile was defined as the minimum conduc) and tivity required at the periphery of the SAN region ( . was denoted as In determining the critical conductivity, each configuration of the model was simulated for a short duration to stabilise any initial transients. At the conductivities required for successful activation of the atrial tissue, all model configurations approached steady-state within 5 s. Each configuration was then simulated for a further 10 s during which the activation state of the atrial tissue region was assessed. A given configuration was deemed to successfully activate the atrial tissue if a 1:1 correspondence was observed between APs in the SAN and atrial regions throughout the entire 10 s observation period. The critical conductivities and ) were then compared to assess the relative ( ease with which each model of SAN heterogeneity could drive the atrial tissue.
Fig. 4. Simulated AP waveshapes for the (a) central and (b) peripheral SAN cell models. Experimental recordings from central and peripheral regions of the SAN to which the model parameters were fitted are also shown (Experimental data from Kodama et al. [15]). In each case, the AP waveshapes are shown aligned at the moment of maximum UV and an enlarged view of a portion of the waveshape (indicated by a dotted rectangle) is shown inset in the upper right corner.
G. Computational Methods The simulation code, including the underlying ionic models was implemented in ANSI C. The ordinary differential equations of the underlying ionic models were evolved using the CVODE solver for stiff systems [14] with a relative error toland a maximum time step of 0.1 ms. The 1-D erance of monodomain equation was solved using a finite-difference approximation of the partial spatial derivatives with a maximum spatial step of 0.0625 mm. III. RESULTS A. Central and Peripheral SAN Cell Models Fig. 4 illustrates the AP waveshapes for the central and peripheral cell models together with the AP data from central and
168
IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 53, NO. 2, FEBRUARY 2006
TABLE I CRITICAL CONDUCTIVITY (S 1 mm
peripheral regions of the SAN to which the default model parameters were fitted [15]. In each case the simulated APs and the experimental recordings are shown aligned at the time of maximum UV. As can be seen, the correlation between the experimental recordings and the simulated AP waveshapes of the central and peripheral cell models is very high.
) FOR ENTRAINMENT
TABLE II CRITICAL CONDUCTIVITY (S 1 mm
) FOR ACTIVATION
B. Critical Conductivity for Entrainment Table I summarizes the critical SAN conductivity for entrainment for each of the SAN models under both the uniform ( ) and the linear ( ) conductivity profiles. For each profile, values are given both in the absence and in the presence of the atrial load. In the case of the mosaic models, the critical conductivity ( ). A number of trials is reported as the of the Mosaic-CA model were unable to maintain spontaneous pacemaker activity. These trials have been omitted in the results summarized in Table I. The number of trials included in the affected results are indicated in the table notes. From Table I it is apparent that the Gradient model achieves frequency entrainment more easily, i.e., at a lower critical conductivity, than the Discrete-Region model. This is true under both the uniform and linear conductivity profiles in the absence and in the presence of the atrial load. Under the uniform conductivity profile, the Gradient model required significantly lower conductivity to achieve entrainment than the Mosaic-PA, Mosaic-CP or Mosaic-CPA models ( ). This was true both in the presence and in the absence of the atrial load. Under the linear conductivity profile, the Gradient model also required a significantly lower conductivity to achieve entrainment in the presence of the atrial load than either ). the Mosaic-PA, Mosaic-CP or Mosaic-CPA models ( This was also true in the absence of the atrial load although in the case of the Mosaic-PA model, the difference was marginally ). less significant ( Under the uniform conductivity profile, the critical conductivity required for entrainment of the Discrete-Region and Mosaic-CP models was unaffected by the addition of the atrial load. The presence of the atrial load also had no significant effect on the critical conductivity required for either the Mosaic-PA or the Mosaic-CPA models. In contrast, the critical conductivity for the Gradient model under the uniform conductivity profile was reduced by approximately 10% on addition of the atrial load. It should be noted that in the case of the Mosaic-CA model, most of the random cell type distributions were unable to maintain
spontaneous pacemaker activity under the uniform conductivity profile (see Table I). Under the linear conductivity profile, addition of the atrial load again had no significant effect on the critical conductivity of either the Discrete-Region, Mosaic-PA or Mosaic-CPA models. However, the critical conductivity for the Mosaic-CP model was significantly increased on addition of the atrial ). As for the uniform conductivity profile, the load ( critical conductivity for the Gradient model under the linear conductivity profile was reduced by approximately 20% on addition of the atrial load. Again, it should be noted that in the case of the Mosaic-CA model, not all of the random cell type distributions were able to maintain spontaneous pacemaker activity (see Table I). C. Critical Conductivity for Activation Table II summarizes the critical conductivity required for activation of the atrial tissue for the different models of SAN heterogeneity, under both the uniform and the linear conductivity profiles. As described above, for the linear conductivity profile, values are given for the critical conductivity at the periphery ) corresponding to a value for the of the SAN region ( ) of 0.01 conductivity at the center of the SAN region ( . Again, in the case of the mosaic models, the critical ( ). conductivity is reported as the The critical conductivity for activation in the case of the Mosaic-CP model under the linear conductivity profile was signif) than that required under the uniicantly lower ( form conductivity profile. However, it was significantly larger ) than that required by either the Discrete-Region or ( Gradient models. The critical conductivities for activation for the Discrete-Region and Gradient models, under the linear conlower than those required ductivity profile were also under the uniform conductivity profile. There was no significant difference in the critical conductivity required for activation of
CLOHERTY et al.: A COMPARISON OF 1-D MODELS OF CARDIAC PACEMAKER HETEROGENEITY
169
Fig. 5. Simulated AP waveshapes for the Discrete-Region model of SAN heterogeneity, (a) and (c) in the absence and (b) and (d) in the presence of the atrial tissue load with the uniform [(a) and (b)] and linear [(c) and (d)] conductivity profiles. In each panel, the upper most trace corresponds to the center of the SAN and the moment of earliest activation has been aligned with t = 0:05 s.
the atrial tissue under the uniform and linear conductivity profiles for either the Mosaic-PA or Mosaic-CPA models. D. Pacemaker Activity in the 1-D SAN Models Figs. 5–9 illustrate the AP waveshapes for each of the models of SAN heterogeneity in the absence (left) and in the presence (right) of the atrial load, for both the uniform (top) and linear (bottom)
conductivity profiles. In the case of the uniform conductivity profile, the SAN conductivity ( ) was set to 25 . Under the linear conductivity profile, the conductivity was set to 0.01 at the center of the SAN ( ) and 25 at the periphery ( ). These values for the conductivity were sufficient to achieve entrainment of the SAN region. In the presence of the atrial load, these values were also sufficient to drive the atrial tissue in all cases except the Mosaic-CA model. In the later
170
IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 53, NO. 2, FEBRUARY 2006
Fig. 6. Simulated AP waveshapes for the Gradient model of SAN heterogeneity, in the absence (a) and (c) and in the presence (b) and (d) of the atrial tissue load with the uniform [(a) and (b)] and linear [(c) and (d)] conductivity profiles. In each panel, the upper most trace corresponds to the center of the SAN and the moment of earliest activation has been aligned with t = 0:05 s.
case, activation of the atrial tissue was not observed at any value of the SAN conductivity with the uniform conductivity profile. Furthermore, these values for the conductivity yield a CV in the SAN , comparable to that observed in region on the order of 0.1 the intact SAN of the rabbit [1]. Curiously, we see in Fig. 5 that even in the absence of the atrial load, the Discrete-Region model exhibits a point of earliest activation not at the periphery of the SAN but partway between the
center and the periphery. This is despite the longer intrinsic CL of the central type cell model (0.477 s) compared to that of the peripheral type cellmodel (0.404 s). This behavior is due to the lower MDP of the central type cell model, which imposes a depolarizing influence on the peripheral type cells, accelerating diastolic depolarization of these peripheral cells and shortening the CL. The entrained CL of the Discrete-Region model with the uni) in the absence of form conductivity profile (
CLOHERTY et al.: A COMPARISON OF 1-D MODELS OF CARDIAC PACEMAKER HETEROGENEITY
171
Fig. 7. Simulated AP waveshapes for the Mosaic-PA model of SAN heterogeneity, (a) and (c) in the absence and (b) and (d) in the presence of the atrial tissue load with the uniform [(a) and (b)] and linear [(c) and (d)] conductivity profiles. In each panel, the upper most trace corresponds to the center of the SAN and the moment of earliest activation has been aligned with t = 0:05 s. The corresponding random cell type assignment for the computational nodes within the SAN region is shown inset, P = Peripheral SAN and A = Atrial type cell models.
the atrial load was 0.381 s, shorter than the intrinsic CL of both the central and peripheral type cell models. This shortening of the entrained CL was also observed in the Mosaic-CP (0.362 s) and Mosaic-CPA (0.396 s) models (Figs. 8 and 9). The entrained CL for the Gradient and Mosaic-PA models shown in Figs. 6 and 7 were 0.412 s and 0.432 s respectively. Fig. 10 shows the CV as a function of position for each of the SAN models under both the uniform and linear conductivity
profiles. In all models of SAN heterogeneity, under both the uniform and linear conductivity profiles, the CV in the SAN region . Under the uniform conducwas on the order of 0.1 tivity profile, both the Gradient and Mosaic-CPA models exhibit , indicating that the AP is inia peak in the CV near tiated almost simultaneously in a small region of tissue rather than at a single point. This peak in the CV was not observed with the linear conductivity profile. Both the Mosaic-PA and
172
IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 53, NO. 2, FEBRUARY 2006
Fig. 8. Simulated AP waveshapes for the Mosaic-CP model of SAN heterogeneity, in the absence (a) and (c) and in the presence (b) and (d) of the atrial tissue load with the uniform [(a) and (b)] and linear [(c) and (d)] conductivity profiles. In each panel, the upper most trace corresponds to the center of the SAN and the moment of earliest activation has been aligned with t = 0:05 s. The corresponding random cell type assignment for the computational nodes within the SAN region is shown inset, C = Central SAN and P = Peripheral SAN type cell models.
Mosaic-CPA models displayed a gradual increase in CV from the center to the periphery of the SAN under the linear conductivity profile. It is interesting to note the marginally higher atrial CV, near the SAN–Atrial border, observed for the Discrete-Region, Gradient and Mosaic-CP models of SAN heterogeneity. In each of these configurations, the atrial tissue is driven by peripheral SAN type cells in the periphery of the SAN region. The periph-
eral SAN cells possess a longer APD than the adjacent atrial tissue and, therefore, supply additional depolarizing current to the atrial tissue region, resulting in a marginally higher CV. The increase in CV occurs only in the proximity of the SAN. In simulations involving an extended atrial tissue region (not shown), the CV in the atrial tissue decreased with distance from the SAN-Atrial interface, stabilizing at the intrinsic atrial value of within approximately 3 mm.
CLOHERTY et al.: A COMPARISON OF 1-D MODELS OF CARDIAC PACEMAKER HETEROGENEITY
173
Fig. 9. Simulated AP waveshapes for the Mosaic-CPA model of SAN heterogeneity, in the absence (a) and (c) and in the presence (b) and (d) of the atrial tissue load with the uniform [(a) and (b)] and linear [(c) and (d)] conductivity profiles. In each panel, the upper most trace corresponds to the center of the SAN and the moment of earliest activation has been aligned with t = 0:05 s. The corresponding random cell type assignment for the computational nodes within the SAN region is shown inset, C = Central SAN, P = Peripheral SAN and A = Atrial type cell models.
IV. DISCUSSION As noted in the Section I, the primary aim of this paper is to identify which of the discrete-region, gradient or mosaic models best accounts for the heterogeneity observed in the SAN. In brief, all of the models except the Mosaic-CA model were able to achieve entrainment of the SAN and successfully excite the adjoining atrial tissue. However, the Mosaic-PA model was un-
able to reproduce the heterogeneity in AP waveshapes characteristic of the SAN. While both the Gradient and Mosaic-CP models produced a site of earliest activation consistent with observations in the SAN, the Gradient model was found to achieve frequency entrainment most easily. Furthermore, the uniform conductivity profile was found to promote entrainment within the SAN while the linear conductivity profile tended to assist propagation of the AP into the atrial tissue. These key points
174
IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 53, NO. 2, FEBRUARY 2006
Fig. 10. Conduction velocity as a function of position for the different models of SAN heterogeneity. (a) under the uniform conductivity profile ( = 25:0 S 1 mm ), and (b) under the linear conductivity profile = 0:01 S 1 mm and = 25:0 S 1 mm ). Note the ( marginally higher atrial CV observed for the Discrete-Region, Gradient and Mosaic-CP models of SAN heterogeneity. The data shown here correspond to the last AP in the observation period. In the case of the mosaic models, the mean conduction velocity as a function of position is shown (n = 10).
are developed in more detail below, followed by a brief discussion of the limitations of this paper. A. SAN–Atrial Conductivity Profiles In perhaps the earliest modeling study aimed at investigating the relationship between the structure and function of the SAN, Joyner and van Capelle [16] constructed a radially symmetric two-dimensional (2-D) model of the SAN and adjoining atrial tissue, somewhat analogous to the discrete-region model of SAN organization described above. They employed a single SAN cell type in the SAN region of their model, coupled to a region of atrial type tissue. In this simple approximation of the discrete-region model they investigated the effect of SAN size and SAN–Atrial coupling on the ability of the SAN to exhibit spontaneous pacemaker activity and also to drive the atrial tissue. Joyner and van Capelle reported that successful excitation of
the atrial tissue required partial uncoupling of the SAN from the atrial tissue, either through a discrete “barrier” resistance or through a region of gradually increasing conductivity. Winslow et al. [17] adjusted the parameters of the NobleDiFrancesco-Denyer [18] model of a single SAN myocyte to produce AP waveshapes similar to those observed in the center and periphery of the SAN by Kodama and Boyett [1]. They employed these modified cell models in a large scale 2-D network model of the SAN with a smooth variation in intrinsic CL from the center to the periphery. In this simple gradient model, the AP was initiated in the periphery of the SAN and propagated inwards to the center. When embedded in a larger atrial network, the site of initial activation shifted from the periphery toward the center of the SAN as observed experimentally by Kirchhof et al. [13]. The Winslow et al. model assumed a common uniform conductivity throughout both SAN and atrial tissue regions and although the SAN was able to successfully excite the atrial tissue, the model was unable to reproduce the higher conduction velocity observed experimentally in atrial tissue compared to that in the SAN. Zhang et al. [19] developed new mathematical models of central and peripheral SAN myocytes and formulated a 1-D gradient model of the SAN. Garny et al. [20] recently published a number of changes to Zhang’s original 1-D gradient model. In this corrected 1-D model of the SAN, APs were initiated in the periphery of the SAN region in the absence of the atrial tissue load. However, with its default parameters, the model was unable to reproduce the shift in the site of earliest activation from the periphery to the center of the SAN region in the presence of an atrial load. It should be noted, however, that Garny et al. used conductivity values based on gap-junction experiments in rabbit SAN and atrial cell pairs [21], [22]. This may underestimate the conductivity in-vivo, since cells communicate with more than one neighbor. In this paper, we have employed two different conductivity profiles within the SAN region of our 1-D models. The first assumes a uniform conductivity throughout the SAN region and is consistent with electron microscopic studies of gap junction density in the SAN and the adjoining atrial tissue. Masson-Pévet et al. [12] estimated the density of gap junctions in the SAN to be roughly 10 times lower than that in the adjacent atrial tissue. The second conductivity profile assumes a linear increase in conductivity from the center to the periphery of the SAN region. A gradient in coupling was first hypothesized by Joyner and van Capelle, and a gradient similar to that employed in this paper was recently employed in the 1-D model formulated by Garny et al. [20]. Though the underlying mechanism is not yet clear, there exists sufficient experimental evidence to surmise such a gradient in conductivity [11]. Both the uniform and linear conductivity profiles employed in this paper include a step change in conductivity at the SAN–Atrial boundary, consistent with the experimental observation of Oosthoek et al. [23] who reported an abrupt increase in gap junction density at the periphery of the SAN. We found that for each of the different models of SAN heterogeneity, the linear conductivity profile resulted in greater dispersion of activation time and marginally lower CV in the SAN. Based on our results in Tables I and II, we note that the
CLOHERTY et al.: A COMPARISON OF 1-D MODELS OF CARDIAC PACEMAKER HETEROGENEITY
uniform conductivity profile favors frequency entrainment of SAN myocytes while the linear conductivity profile marginally favors propagation of the AP from the SAN into the atrial tissue. This is broadly consistent with the results of the modeling study by Joyner and van Capelle [16] who reported that partial uncoupling of the SAN from the atrial tissue, either through a discrete “barrier” resistance or through a region of gradually increasing conductivity, was required for successful excitation of the atrial tissue. Joyner and van Capelle also noted that a gradual transition in conductivity between the SAN and atrial regions allowed a smaller SAN region to successfully drive the atria than was the case for a discrete barrier resistance. Although we report qualitatively similar results, we note that in this paper, the atria could be successfully driven even with the uniform conductivity profile. B. Pacemaker Activity Similar to the modeling study by Joyner and van Capelle [16], we observed three states of activity in which the SAN was either: completely quiescent, spontaneously active but unable to excite the atria, or spontaneously active and able to successfully excite the atrial tissue. The Mosaic-CA model involving central SAN and atrial type cell models was found to be untenable, largely failing to maintain spontaneous activity and drive the atrial tissue. Under the uniform conductivity profile, at low vales of , the central SAN cells fired spontaneously, but were unable to activate the was increased, atrial cells in interspersed atrial cells. As the central portion of the SAN, where the proportion of such cells was relatively low, were stimulated to threshold, eliciting APs. However, automaticity of the cells in the periphery of the SAN, where the proportion of atrial cells was considerably was increased. With higher, was increasingly suppressed as the uniform conductivity profile, only 2 of the 10 random cell type distributions of the Mosaic-CA model were capable of entrained pacemaker activity. We found that the Discrete-Region, Gradient, Mosaic-PA, Mosaic-CP and Mosaic-CPA models were all able to achieve coordinated activation of the entire SAN region and initiate APs in the adjoining atrial tissue region as shown in Figs. 5–9. In the mosaic models, peripheral SAN myocytes were found to be better suited to driving the interspersed atrial cells than the central SAN myocytes. A similar result was reported by Zhang et al. when using their central and peripheral SAN cell models together with an atrial cell model to formulate mosaic models of the SAN [24]. C. Site of Earliest Activation In both the Gradient and Mosaic-CP models, under both the uniform and linear conductivity profiles, the site of earliest activation was located at the SAN-Atrial boundary in the absence of the atrial load. In the presence of the atrial load, the site of earliest activation was shifted away from the periphery and toward the center of the SAN. This is consistent with the experimental observations by Kirchoff et al. [13]. In contrast, the site of earliest activation in the Mosaic-PA model was located at the center of the SAN under both the uniform and linear conductivity profiles and was unaffected by the atrial load.
175
As noted in the results, for the Discrete-Region and Mosaic-CPA models, the more positive MDP of the central SAN cells has a depolarizing effect on adjacent peripheral SAN cells. This accelerates the diastolic depolarization of the peripheral cells and shortens the entrained CL. As a result, the site of initial activation lies at an intermediate point between the center and periphery of the SAN region. The shortening of the entrained CL in the 1-D model is in contrast to the entrained CL observed when a single central SAN cell model is coupled to a single peripheral SAN cell model, where the entrained CL lies somewhere between the two intrinsic CLs [2]. Activation maps from the intact SAN suggest that the site of earliest activation lies in a region of SAN tissue with intermediate AP waveshape characteristics [25]. This is consistent with the behavior observed in the Discrete-Region, Gradient, Mosaic-CP and Mosaic-CPA models of this paper. We also observed slow conduction from the primary pacemaker site back ) of the SAN (Fig. 10). This is compatoward the center ( rable to the slow conduction observed experimentally in the intact SAN in the direction of the interatrial septum [25]. It, therefore, seems plausible that the characteristic spread of excitation from the SAN may at least in part be due to the regional distribution of cell types in and around the SAN. This is consistent with the view that the block zone on the septal side of the SAN is due to a region of cells exhibiting low excitability [9], [25]. It also suggests that central SAN cells in or adjacent to the block zone may play an important role in normal pacemaker function by acting as a current source supplying current to the primary pacemaker site. D. Preservation of Action Potential Heterogeneity Kirchhof et al. [13] reported that dissecting away the atrial tissue from the SAN resulted in a decrease in CL and a shift of the dominant pacemaking site away from the center and toward the periphery of the SAN. This appeared to be at odds with observations made in the intact SAN and provided evidence that the observed AP waveshape of a cell is determined not only by the intrinsic dynamics of the cell membrane, but is also at least partially influenced by the electrotonic interaction with neighboring cells. In the discrete-region model of the SAN, the observed heterogeneity in AP waveshape was, therefore, attributed to the electrotonic load imposed on the SAN by the surrounding more hyperpolarized atrial tissue. Though not the primary focus of that study, the discrete-region model of Joyner and van Capelle [16] appeared to exhibit a central-peripheral variation in SAN AP waveshapes at least qualitatively consistent with that observed in the intact SAN. This variation in AP waveshape was due entirely to the electrotonic influence of the quiescent atrial tissue region. Using their models of central and peripheral SAN myocytes, Zhang et al. [24] also formulated 2-D mosaic models of central and peripheral SAN tissue. They randomly assigned either SAN or atrial cell membrane properties to each node in the central and peripheral network models according to the proportions reported by Verheijck et al. [4], although they incorrectly assigned a higher proportion of atrial cells in the center of the node (41%) compared with the data of Verheijck et al. (22%). Zhang et al.
176
IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 53, NO. 2, FEBRUARY 2006
concluded that the mosaic model of the SAN could not account for the observed heterogeneity in AP characteristics [24]. In this paper, we have employed single SAN cell models capable of reproducing in isolation the experimental AP recordings from the center and periphery of the SAN. The Discrete-Region, Gradient, Mosaic-CP, and Mosaic-CPA models all displayed AP waveshapes characteristic of both the center and periphery of the SAN. However, the Mosaic-PA variation of the mosaic model contained no central SAN cells and was, therefore, unable to produce AP waveshapes characteristic of the central region of the SAN. E. Entrainment of the SAN and Activation of the Atrial Tissue The Gradient model of SAN heterogeneity was found to achieve frequency entrainment of the SAN considerably more easily than either the Discrete-Region or the mosaic models. However, in terms of the critical conductivity for activation of the atrial tissue region, there was little to distinguish between the different models. This was true for both the uniform and the linear conductivity profiles, although the linear conductivity profile slightly favored activation of the atrial tissue. This suggests that successful propagation of the AP from the SAN into the atrial tissue is largely independent of the nature of the heterogeneity within the SAN. Considering all of the results together, the Gradient model of SAN heterogeneity, with the uniform conductivity profile, was most able to achieve frequency entrainment and successfully drive the atrial tissue. F. Limitations of the Present Study Based on the results presented above, and the points already noted in the discussion, it seems that the Gradient model with the uniform conductivity profile represents the most plausible model of heterogeneity in the SAN. Yet, conventional wisdom suggests that the site of earliest activation in the presence of the atrial load [see Fig. 6(b)] should lie at the center of the SAN region, i.e., at . The apparent off-center site of pacemaker initiation in Fig. 6(b) may be reconciled by noting that the upstroke velocity of the central cells is slower than cells off-center. It is difficult to compare activation times in cells with markedly differing upstroke rates, in order to determine which is activated first. The important observation, as noted above, is that the site of initial activation shifts toward the center in the presence of the atrial tissue load, in accordance with experimental observations. In the simulations involving the linear conductivity profile, ), is lower the conductivity at the center of the SAN region ( than might be expected in the SAN. However, the value for was chosen to facilitate differentiation between the different models based on their ability to achieve frequency entrainment under the linear conductivity profile. The AP waveshapes illustrated in Figs. 5–9 for the linear conductivity profile, therefore, represent an extreme of the behavior expected of the different models in the presence of a gradient in conduc, even to tivity. As can be seen in these figures, increasing the value corresponding to the uniform conductivity profile, has only a marginal effect on both the site of earliest activation and the observed AP waveshapes.
In their study of gap junction distribution in the SAN, Oosthoek et al. [23] observed small bundles of nodal cells penetrating the atrial myocardium. Inspired by this tissue architecture, Noble and Winslow [26] formulated a 2-D network model of the rabbit SAN and surrounding atrial tissue. This model incorporated fingers of SAN tissue extending from the periphery of the SAN into the surrounding atrial tissue region. They demonstrated that the interdigitation of SAN and atrial tissue provided a viable mechanism to prevent suppression of spontaneous activity in the SAN and facilitate propagation of the activation wavefront from the SAN into the surrounding atrial tissue. The results of Noble and Winslow highlighted the potential importance of tissue geometry in facilitating propagation of the AP from the SAN into the atrial tissue. In the 1-D models described above, it was not possible to model different geometric configurations. Nevertheless, we were able to achieve activation of the atrial tissue, in the presence and absence of a gradient in conductivity, in all but the Mosaic-CA model. Furthermore, a 1-D model may serve as a reasonable approximation of a radially symmetric 2-D model of comparable dimension [16], [19]. Therefore, given the reduced computational demands compared to higher dimensional models, the 1-D model described here provides a viable tool for the investigation of SAN function. The use of a 1-D model in this paper also facilitates a direct comparison with 1-D modeling studies reported by others (e.g., [19] and [20]).
V. CONCLUSION We have previously demonstrated that heterogeneity in AP waveshape assists frequency entrainment of electrically coupled pacemaker cell pairs [2]. In this paper we have investigated a number of different conceptual models of SAN heterogeneity using detailed ionic models of electrical activity in SAN and atrial myocytes. To our knowledge, this represents the first attempt to formulate and compare each of the different models of SAN heterogeneity. A total of six different 1-D models of SAN heterogeneity were formulated, namely the Discrete-Region, Gradient, Mosaic-CA, Mosaic-PA, Mosaic-CP and Mosaic-CPA models. Each of the different models were assessed on their ability to achieve frequency entrainment of the SAN and activation of the adjoining atrial tissue. We observed that both the Gradient and Mosaic-CP models displayed AP waveshapes characteristic of central and peripheral SAN myocytes, and displayed a shift in the site of earliest activation from the periphery toward the center of the SAN in the presence of the atrial load. Both the Gradient and Mosaic-CP models would, therefore, appear to produce behavior consistent with experimental observations in the intact SAN. The Gradient model achieved frequency entrainment considerably more easily than the other models of SAN heterogeneity, including the Mosaic-CP model. In contrast, SAN heterogeneity appears to play only a minor role in facilitating propagation of the AP from the SAN into the atrial tissue. We observed that a uniform conductivity throughout the SAN significantly promoted frequency entrainment of the SAN while
CLOHERTY et al.: A COMPARISON OF 1-D MODELS OF CARDIAC PACEMAKER HETEROGENEITY
a gradual increase in conductivity from the center to the periphery of the SAN marginally promoted propagation of the AP from the SAN into the atrial tissue. Based on these results, the gradient model of SAN heterogeneity with a uniform SAN conductivity is the most likely model of SAN organization, able to qualitatively reproduce known behavior of the intact cardiac pacemaker. REFERENCES [1] I. Kodama and M. R. Boyett, “Regional differences in the electrical activity of the rabbit sinus node,” Pflügers Arch., vol. 404, pp. 214–226, 1985. [2] S. L. Cloherty, N. H. Lovell, B. G. Celler, and S. Dokos, “Inhomogeneity in action potential waveshape assists frequency entrainment of cardiac pacemaker cells,” IEEE Trans. Biomed. Eng., vol. 48, pp. 1108–1115, Oct. 2001. [3] L. N. Bouman and H. J. Jongsma, “Structure and function of the sinoatrial node: A review,” Eur. Heart J., vol. 7, pp. 94–104, 1986. [4] E. E. Verheijck, A. Wessels, A. C. G. van Ginneken, J. Bourier, M. W. M. Markman, J. L. M. Vermeulen, J. M. T. de Bakker, W. H. Lamers, T. Opthof, and L. N. Bouman, “Distribution of atrial and nodal cells within the rabbit sinoatrial node,” Circulation, vol. 97, pp. 1623–1631, 1998. [5] N. H. Lovell, S. L. Cloherty, B. G. Celler, and S. Dokos, “A gradient model of cardiac pacemaker myocytes,” Prog. Biophys. Mol. Biol., vol. 85, no. 2–3, pp. 301–323, Jun.–Jul. 2004. [6] S. Dokos and N. H. Lovell, “A curvilinear gradient path method for optimization of biological systems models,” in Proc. 5th IFAC Symp. Modeling and Control in Biological Systems, D. D. Feng and E. R. Carson, Eds., Oxford, U.K., 2003, pp. 203–208. [7] Y. E. Earm and D. Noble, “A model of the single atrial cell: Relation between calcium current and calcium release,” Proc. R. Soc. Lond. B, vol. 240, pp. 83–96, 1990. [8] D. W. Hilgemann and D. Noble, “Excitation-contraction coupling and extracellular calcium transients in rabbit atrium: Reconstruction of basic cellular mechanisms,” Proc. R. Soc. Lond. B, vol. 230, pp. 163–205, 1987. [9] W. K. Bleeker, A. J. C. MacKaay, M. Masson-Pévet, L. N. Bouman, and A. E. Becker, “Functional and morphological organization of the rabbit sinus node,” Circ. Res., vol. 46, no. 1, pp. 11–22, Jan. 1980. [10] F. I. M. Bonke, “Passive electrical properties of atrial fibers of the rabbit heart,” Pflügers Arch., vol. 339, pp. 1–15, 1973. [11] M. R. Boyett, H. Honjo, and I. Kodama, “The sinoatrial node, A heterogeneous pacemaker structure,” Cardiovasc. Res., vol. 47, pp. 658–687, 2000. [12] M. Masson-Pévet, W. K. Bleeker, and D. Gros, “The plasma membrane of leading pacemaker cells in the rabbit sinus node: A qualitative and quantitative ultrastructural analysis,” Circ. Res., vol. 45, pp. 621–629, 1979. [13] C. J. H. J. Kirchhof, F. I. M. Bonke, M. A. Allessie, and W. J. E. P. Lammers, “The influence of the atrial myocardium on impulse formation in the rabbit sinus node,” Pflügers Arch., vol. 410, pp. 198–203, 1987. [14] S. Cohen and A. Hindmarsh, “CVODE, A stiff/nonstiff ODE solver in C,” Comput. Phys., vol. 10, no. 2, pp. 138–143, 1996. [15] I. Kodama, M. R. Boyett, M. R. Nikmaram, M. Yamamoto, H. Honjo, and R. Niwa, “Regional differences in effects of E-4031 within the sinoatrial node,” Am. J. Physiol., vol. 276, pp. H793–H802, 1999. [16] R. W. Joyner and F. J. L. van Capelle, “Propagation through electrically coupled cells: How a small SA node drives a large atrium,” Biophys. J., vol. 50, pp. 1157–1164, Dec. 1986. [17] R. L. Winslow, D. Cai, A. Varghese, and Y. Lai, “Generation and propagation of normal and abnormal pacemaker activity in network models of cardiac sinus node and atrium,” Chaos, Solitons, Fractals, vol. 5, no. 3/4, pp. 491–512, 1995. [18] D. Noble, D. DiFrancesco, and J. Denyer, “Ionic mechanisms in normal and abnormal cardiac pacemaker activity,” in Neuronal and Cellular Oscillators, J. W. Jacklet, Ed. New York: Marcel Dekker, 1989, pp. 59–85. [19] H. Zhang, A. V. Holden, I. Kodama, H. Honjo, M. Lei, T. Varghese, and M. R. Boyett, “Mathematical models of action potentials in the periphery and center of the rabbit sinoatrial node,” Am. J. Physiol., vol. 279, pp. H397–H421, 2000. [20] A. Garny, P. Kohl, P. J. Hunter, M. R. Boyett, and D. Noble, “One-dimensional rabbit sinoatrial node models: Benefits and limitations,” J. Cardiovasc. Electrophysiol., vol. 14, no. s10, pp. S121–S121, 2003.
177
[21] S. Verheule, M. J. A. van Kempen, S. Postma, M. B. Rook, and H. J. Jongsma, “Gap junctions in the rabbit sinoatrial node,” Am. J. Physiol., vol. 280, pp. H2103–H2115, 2001. [22] S. Verheule, M. J. A. van Kempen, P. H. J. A. te Welscher, B. R. Kwak, and H. J. Jongsma, “Characterization of gap junction channels in adult rabbit atrial and ventricular myocardium,” Circ. Res., vol. 80, no. 5, pp. 673–681, 1997. [23] P. W. Oosthoek, S. Virágh, A. E. M. Mayen, M. J. A. van Kempen, W. H. Lamers, and A. F. M. Moorman, “Immunohistochemical delineation of the conduction system I: The sinoatrial node,” Circ. Res., vol. 73, pp. 473–481, 1993. [24] H. Zhang, A. V. Holden, and M. R. Boyett, “Gradient model versus mosaic model of the sinoatrial node,” Circulation, vol. 103, pp. 584–588, 2001. [25] M. R. Boyett, H. Honjo, M. Yamamoto, M. R. Nikmaram, R. Niwa, and I. Kodama, “Downward gradient in action potential duration along the conduction path in and around the sinoatrial node,” Am. J. Physiol., vol. 276, pp. H686–H698, 1999. [26] D. Noble and R. L. Winslow, “Reconstructing the heart: Network models of SA node–atrial interaction,” in Computational Biology of the Heart, A. V. Panfilov and A. V. Holden, Eds. Chichester, U.K.: Wiley , 1997, pp. 49–64.
Shaun L. Cloherty (S’01–M’05) received the B.E. (Hons) degree in aerospace avionics from the Queensland University of Technology, Brisbane, Australia, in 1997. He received the Ph.D. degree in biomedical engineering from the University of New South Wales, Sydney, Australia, in 2005 under the supervision of Prof. N. H. Lovell and Dr. S. Dokos. He is currently a Research Associate with the Graduate School of Biomedical Engineering at the University of New South Wales, Sydney, Australia. His current research interests include cardiac electrophysiology and modeling, flow estimation and modeling for control of an implantable rotary blood pump, and modeling of electrical stimulation strategies for selective excitation of neural tissue for a visual prosthesis.
Socrates Dokos received the B.E. (Hons) degree in electrical engineering in 1987 and the Ph.D. degree in biomedical engineering in 1996 from the University of New South Wales, Sydney, Australia. From 1997–1999, he held a postdoctoral position in the department of Physiology at Auckland University School of Medicine, Auckland, New Zealand, where he investigated mechanical properties of heart tissue. He is currently a Lecturer with the Graduate School of Biomedical Engineering at the University of New South Wales. His research interests include ionic modeling of excitable cardiac, neural, and retinal tissues. He is also pursuing large-scale parameter optimization techniques for developing biophysically accurate models of excitable cells.
Nigel H. Lovell (M’91–SM’99) received the B.E. (Hons) and Ph.D. degrees from the University of New South Wales, Sydney, Australia. He is currently a Professor in the Graduate School of Biomedical Engineering, University of New South Wales. His research work has covered areas of expertise ranging from cardiac modeling, home telecare technologies, biological signal processing and visual prosthesis design, having authored over 150 refereed journals, conference proceedings, book chapters, and patents. Dr. Lovell is currently the IEEE Engineering in Medicine and Biology Society (EMBS) Vice President (VP) for Conferences (2004–2005) and was the VP for Member and Student Activities (2002–2003). He was awarded the IEEE Third Millennium Medal for services to the EMBS.
