Simulated reality

by Elizabeth Murray

Researchers at the University of Illinois take on a challenging fluid mechanics problem to model blood flow in the cardiovascular system to improve clinical diagnostic tools.

It may be hard to imagine, but you can live without a heartbeat. Though as you may have guessed, medically speaking, it is not ideal. Patients with terminal heart failures, often times while waiting for a donor heart, are put on ventricular assistive devices such as artificial hearts.

Artificial hearts work as axial or centrifugal pumps, carefully designed to pass six to ten liters of blood per minute, a rate that is close to the flow from a normal heart. Axial flow devices have a spiral propeller that takes in blood and pushes it forward continuously. Arif Masud, a professor in the Department of Civil and Environmental Engineering at the University of Illinois at Urbana-Champaign, says to think of it like a fan in a room; as long as it is running, air is continuously blowing in a uniform fashion.

“But it’s more complicated than that because blood is not like air or water,” he clarifies. “Studying blood flow isn’t like our typical fluid mechanics research; it doesn’t flow like air around an aircraft or water through a pipeline. We are dealing with a far more complex material.”

In the human system, the natural heart continuously beats and stops, beats and stops, beats and stops. As a result, blood flow from a natural heart is not uniform; it pulses. Flow that is coming out of an artificial heart is at a near steady rate, and while the net volume is same as that of a natural heart, the pulsation is gone—the rapid movement of the blood stream starting and stopping is lost. This creates a standing wave in the flow, which is why the patient no longer has a heartbeat.

This lack of heartbeat does not cause the patient any immediate danger; in fact, as long as the patient is receiving the same amount of blood to their vital organs, it would seem they would continue to survive without an incident. However, for many of these patients a very real problem develops months later in the form of a stroke.

So how is it that clot formation takes place in flowing blood?

“If you look at blood under a microscope you’ll see it is a complex mixture of cells and plasma. It has both red blood cells and white blood cells,” Masud explains. “If you don’t shake them, the white platelets have a tendency to coagulate.” Most of the time the beating of the natural heart provides enough pulsation to shake those platelets loose and avoid clotting.

For Masud and his team—who specialize in the mathematical and computational research aspects of structural and fluid mechanics—it’s not about changing the way in which the pump is made; they have actually shied away from getting involved in the designing of a pump. Their objective is to show that the technology exists—or at the very least, that there are people like them who can develop a technology—that can be used for medical diagnosis in much more detail than has been done to date.

“Our goal is to use simulated reality to find the missing pieces of information that can augment MRI or ultrasound data for more comprehensive clinical diagnostics,” says Masud.

Diagnostics in simulated environments

Once Blue Waters came online last year, the team turned to the sustained petascale power of this supercomputer to run their mathematical models and codes for simulating blood flow through diseased arteries. Their objective has been to understand dissection, clot formation, and stroke with the hopes of giving doctors better tools to diagnose the risk factors for patients and treat the problems quickly and more efficiently. Research Associate JaeHyuk Kwack, who worked on the development of blood flow models as the focus of his PhD thesis research at the University of Illinois at Urbana-Champaign, successfully ported their in-house code onto Blue Waters platform to run patient specific models with realistic pulsatile pressures and flow conditions. The team is working with a group of doctors from the Center for Heart Transplant and Assist Devices located in Chicago.

Masud says his team’s new diagnostic technology is similar to the way an X-ray shows the crack in a broken bone, only it would be a simulation based insight for your arteries. It comes down to better, faster, smarter computational methods applied to cardiovascular flows that he calls “simulated reality.” The team carried out flow simulations in the carotid artery in a patient with stenosis in the main artery and aneurysm at the bifurcation.

For now, the risk level of a patient with diseased arteries is not always obvious to the doctors. It isn’t something they can see on an MRI, furthermore ultrasounds and venographs only capture the general flow features.

“It’s what is actually happening in the local fluctuations that gives rise to the causes of the disease, and these simulations and visualizations give us a look inside,” says Masud.

It seems pretty obvious that in a healthy artery, red blood cells flow through smoothly. But in a diseased artery, these cells are stressed as they go through the constrictions.

“Through these simulations, we see that local velocity changes rapidly—often times seen as a jet effect through blockage that can damage the endothelial cell layer—along with causing the red blood cells to spiral around on the inner surface of the artery,” says Masud.

The spiraling of these cells acts as a grinder, not only damaging the cells themselves, but also causing damage to the endothelial layer. This effect is related to what’s called the wall shear stress, which is considered to be the main reason for progression of disease in the arteries. Further, as a direct result of the damage to the endothelial cell layer, the body kicks in its self-defense mechanism and tries to repair the area by sending white platelets that develop a platelet plug. If hemodynamic forces cause this clump of platelets and fibrin to break away, a clot is formed called thrombus that flows down to the narrower arteries and blocks them.

Another major issue is aneurysm or ballooning of the artery wall, which is dangerous not only in its own right as it can rupture under higher pressures, but can also lead to local stagnation in the flow during diastole. This gives white platelets enough time to start sticking together to form a mass of thickened blood, which can make flow through smaller arteries more difficult. Because of the blockade caused by clots or this mass of thickened blood the oxygen demand is not met in the brain cells and they start dying out. “Stroke is inevitable,” says Masud.

“With our new kind of visualizations, we can actually see what doctors know to be true but cannot see,” he continues. “For example, they could help the doctors determine: Should we do a bypass graft? Can we just put in a stent and it will be fine? Or if we have two patients with exactly the same symptoms, this constriction and ballooning effect, which of these two is in more critical condition? What is their risk of dissection, risk of clotting, risk of stroke? And if we have to schedule some procedures for them, who should go in first?”

To begin to visualize the 3D nature of blood flow within these diseased arteries, the team turned to NCSA’s Advanced Digital Services Visualization Group. They worked with visualization programmer Mark Van Moer and used ParaView, a parallel renderer, with custom ffmpeg software to create the first round of visualizations.

New supercomputing paradigm

If you look at what Masud and his team have done computationally speaking in the past—in terms of transient fluid mechanics models, fluid-structure interaction models, techniques that can model vortices and spinning effects, and turbulence—you start to see obvious links to what they did in terms of modeling blood flow, with one key difference: access to a petascale powerhouse.

“Biological systems have a degree of uncertainty associated with them that requires a sequence of these high-end and heavy computations. With Blue Waters coming online, we finally had the compute power,” Masud says. The missing piece of the puzzle was finally found.

For Masud, Blue Waters had the unique hardware architecture his team had been waiting for.

“You have substantial memory at the local processing nodes. Our new computational methods take advantage of this kind of memory to make our models smart, and this helps in minimizing the communication between the processors,” he explains. “What we’re doing is reducing the global communication cost in favor of doing local computations on the resident memory of the processing nodes because these computations are inexpensive. This results in high fidelity simulations at a substantially reduced cost compared to our competing methods.”

Thanks to Blue Waters, the team has more than a proof of concept, they have proven methods that are already showing relevance to the medical question at hand and are finding answers through computer simulations.

“The things that I’m hoping will come out of all of our work—outside of the now obvious successes, which are the new blood flow models that were developed here—is that someday in the near future they’ll have actual practical utility,” says Masud. “I also hope this work will pave the way for a change in the mindset of how we develop mathematical methods for supercomputing moving forward.”

Disclaimer: Due to changes in website systems, we've adjusted archived content to fit the present-day site and the articles will not appear in their original published format. Formatting, header information, photographs and other illustrations are not available in archived articles.