Pursuing an End to Sudden Cardiac Death

nyone who's watched a hospital drama is familiar with the scene: A heart monitor flatlines, accompanied by an eerie alarm signal. A doctor or nurse calls out, "He's in V-fib!" and electric paddles are charged and applied to the patient's chest. The hope is that the shock will correct the rhythm of the heart, saving the patient's life--at least for the moment. UCLA cardiac researcher Alan Garfinkel studies V-fib, short for ventricular fibrillation. Using NPACI's IBM SP system, he and his team are modeling the electrical signals that pulse through the specialized cells of the heart muscle. The simulations help them understand how electrical waves become chaotic in ventricular fibrillation (VF), the major cause of sudden cardiac death.

The heart and the brain are the body's two most critical organs. When either ceases to function, death ensues within minutes. The complexities of the brain are legendary, but the heart--a highly specialized muscle--is still quite intricate. Oxygen-poor blood is received from the body via the superior and inferior vena cava, which feed into the right side of the heart. The blood moves from the right atrium through a valve and into the right ventricle and then is pumped through the right pulmonary artery out to the lungs where carbon dioxide is released and a new supply of oxygen is received. The rich, red blood then travels the left pulmonary artery back to the heart, is pumped from the left atrium through a valve to the left ventricle, and is pumped out to the body via the aorta.

CELLULAR COMMUNICATION

MODEL BEHAVIOR

RESTITUTION

WHOLEHEARTED TERAFLOPS SUPPORT

CELLULAR COMMUNICATION

Coordination of this activity occurs on a cellular level. The contractile tissue of the heart is composed of specialized cells called myocytes, which are joined by other specialized pacemaker cells. With proper frequency, pacemaker cells trigger the release of a rush of ions that are translated into an action potential, or electrical signal. The myocytes are coupled to allow these signals to propagate quickly through the tissue. They travel in more-or-less rectilinear waves, causing the heart to contract and relax. This heartbeat action moves blood through the heart and out to the body.

The ventricles, which are much larger in size than the atria that receive blood, are the most critical chambers of the heart, as they are responsible for generating the thrust to pump blood out to the lungs or to the body. It is possible for a patient to live with atria that do not function efficiently, but if the ventricles stop contracting, as they do in the frenzy of VF, death occurs within minutes.

Garfinkel and his research group at the UCLA Department of Medicine (Cardiology) and Department of Physiological Science are focusing on the normal, planar wave of action potential that travels through the heart, propagating rhythmic contraction. While many pathological conditions--such as extended runs of premature ventricular contractions--are known to contribute to the breakdown of these planar waves and lead to VF, in some instances the normal wave becomes chaotic, seemingly without such an external influence.

The electrical activity of cardiac cells involved in ventricular fibrillation (VF), the leading cause of sudden cardiac death. Left: A single scroll wave has broken down into the chaotic state of VF. Right: A simulated drug has been delivered to the cells, altering their response to premature stimuli. This drug prevents the breakup into VF. In both images, only the cells that are currently firing are shown; the rest of the tissue has been rendered transparent. Images courtesy of Alan Garfinkel, Zhilin Qu, Jong Kil, and James N. Weiss.

MODEL BEHAVIOR

"Supercomputing is absolutely essential to understanding VF," Garfinkel says. "There are very, very basic equations that can be written to describe how a heart cell operates. Complexity is introduced when we jump up to the level of modeling heart tissue, composed of millions of heart cell models."

Many cardiac arrhythmias are time- and space-sensitive and can only be understood in the context of three dimensions. For example, the UCLA group has investigated the phenomenon of spiral wave breakup as the mechanism underlying the genesis of VF. In their scenario, the planar wave of normal conduction first becomes unstable when it confronts diseased or dysfunctional patches of tissue. These slow down part of the wave and cause the remaining part to circulate around the dysfunctional region. Then this re-entrant wave breaks down further into the chaos of VF.

RESTITUTION

"Our theoretical and computer simulation studies have led us to hypothesize that restitution is the key dynamical property responsible for destabilizing spiral waves," Garfinkel says. "When a wave of excitation passes into tissue that is not yet fully recovered from the previous excitation, it results in an incomplete excitation, whose duration is shorter than a full-blown one."

The relationship between the rest period since the last excitation ended and the next action phase duration is called the "restitution curve." The steeper the restitution curve, the more likely the spiral wave of electricity is to break down, causing VF.

After doing several 2-D studies on lab workstations, the UCLA group began using supercomputers to model a cube of heart tissue 320 x 320 x 100 computational cells in size. "The real human heart, realized at full cellular resolution, would require on the order of 10¹⁰ cells," he says. "We have found that the largest space step that avoids instability and provides reasonable accuracy is 0.15 millimeters, which is the basis of the cell size we use in our model."

An allocation on the 144-processor IBM SP system at SDSC--which was recently superseded by Blue Horizon--allowed Garfinkel and his colleagues to simulate both stable and unstable spiral waves in 3-D. "We were able to confirm the role of restitution in causing the chaotic breakdown of electrical signals," Garfinkel says. "From there, we worked backwards, determining how the parameters of the model would need to be altered to lower the slope of the restitution curve." These modeling studies have already led to investigations into drug therapies to treat arrhythmia.

WHOLEHEARTED TERAFLOPS SUPPORT

"Of course, we would be more certain and complete in our findings if we were able to model the whole heart," Garfinkel says. "The gross anatomy of the heart plays such a critical role in the conduction of electrical activity. We're accounting for anatomy in a very small and limited space, and it already takes seven trillion separate mathematical computations to simulate one second of real heart activity."

The simulation also took 22 hours, running on 50 processors of the 144-processor IBM SP system. "We're very enthusiastic about the computing capabilities of the teraflops SP system," Garfinkel says. "We already have the code for a whole-heart model. The only thing preventing us from simulating spiral wave breakdown and VF on a whole heart is processing time. It would take months on the same system we ran our smaller models on."

The UCLA group plans to apply for an allocation on Blue Horizon when it goes into full production status, and to continue their VF studies. "We're also very supportive of the terascale computing initiative recently announced by the National Science Foundation," he says. "Terascale computing will help us understand and eventually prevent and counteract this major cause of sudden cardiac death." --AF *