TACC Supercomputers Render Artificial Heart Valves

Hearts really do take a beating. A human heart beats about 3 billion times in a person’s life. Each pulse puts wear and tear on the constituent cells on the heart valve leaflets, which constantly lose and add new cells in response to the strain of being the heart’s gatekeepers.

Not all hearts are up to the job. Between 75,000 and 100,000 people in North America receive heart valve implants each year, and several hundred thousand people receive them worldwide. The first such replacement valves were made from synthetic materials. These “mechanical” valves last indefinitely but require lifelong treatment with anticoagulants (blood thinners), which requires monthly blood tests to monitor dosage and has associated medical complications.

In the last 20 years, “bioprosthetic” heart valves fabricated from animal tissue sources have become the dominate design, as they do not require the use of anticoagulant drugs because of their superior blood flow dynamics. However, they have one big drawback: they wear out in 10 to 15 years. Doctors generally limit the use of bioprosthetic valves to older patients and rely on older-technology mechanical valves for younger patients, because replacing the valve every decade is impractical.

Professors Michael Sacks and Ming-Chen Hsu are researching ways to create more durable heart valves. Sacks is Director of the Institute for Computational Engineering and Sciences (ICES) Center for Cardiovascular Simulation and professor of biomedical engineering and holder of the W. A. “Tex” Moncrief, Jr. Simulation-Based Engineering Science Chair I at The University of Texas at Austin (UT Austin). Hsu is Assistant Professor in the Department of Mechanical Engineering at Iowa State University. “We’re focusing on two things: creating novel material models for how these tissues respond to being in the body and understanding how the valve works inside a blood flow environment,” says Sacks. “We’re developing new material models that are much more structurally and biologically informed and can actually integrate mechanisms of failure, remodeling, growth, and adaptation to altered forces that go on inside the body.”

Sacks sees this research impacting just about anybody who’s going to get the next generation of heart valve replacements, especially the percutaneous, or non-surgically replaced, artificial valve. “If you could make a valve that would last even three to five years longer, it would have a huge clinical impact and corner the market,” Sacks says. “Our work may also help surgeons develop improved surgical techniques when repairing heart valves.” Hsu and Sacks, along with their collaborators, recently developed a new fluid-structure interaction (FSI) model to predict the long-term performance and durability of an artificial heart valve. Their model simulates not only the valve function but the complex blood flow patterns and how the leaflets behave and change over time in response to that fluid flow.

Stampede is one of the most powerful and significant current supercomputers in the U.S. for open science research. Able to perform nearly 10 quadrillion operations per second, Stampede offers infinite opportunities for computational science and technology. This Dell PowerEdge cluster equipped with Intel Xeon Phi coprocessors pushes the envelope of computational capabilities. It contains 6,400 nodes and 522,000 processing cores.
Stampede is one of the most powerful and significant current supercomputers in the U.S. for open science research. Able to perform nearly 10 quadrillion operations per second, Stampede offers infinite opportunities for computational science and technology. This Dell PowerEdge cluster equipped with Intel Xeon Phi coprocessors pushes the envelope of computational capabilities. It contains 6,400 nodes and 522,000 processing cores.

Supercomputers have been key to running these simulations and making stronger, longer-lasting heart valves. Sacks’ and Hsu’s research team has used millions of CPU hours on Intel® processor-based supercomputers at the Texas Advanced Computing Center (TACC) at UT Austin to understand how heart valve tissue responds to blood flow.

The researchers run countless simulations to predict how the heart valve leaflet behaves and fatigues after, say, a million cycles. It’s a coupled dynamic problem, because the changing blood flow changes how the leaflet behaves, and vice versa.

”The most significant contribution of our current work with supercomputers is our development of an accurate and robust simulation framework that considers the interaction and coupling between hemodynamics and leaflet motions,” Hsu says. “Such an approach involves fluid–structure interaction simulation, which considers a complete mechanical environment of the valve motion and generates more accurate pressure distributions during the dynamic cardiac cycle.”

The research team developed cutting-edge mathematical tools for these simulations that require supercomputer-caliber processing power. “TACC resources are crucial to our research, because the problems are very, very large,” Sacks says. “These jobs run multiple processors and are very computationally intensive. Without them, we simply could not do this work.”

TACC designs and operates some of the world’s most powerful computing resources. The center’s mission is to enable discoveries that advance science and society through the application of advanced computing technologies. TACC’s environment includes resources in high performance computing (HPC), visualization, data analysis, storage, archive, cloud, data-driven computing, connectivity, tools, APIs, algorithms, consulting, and software. TACC works with thousands of researchers on more than 3,000 projects a year.

with 1,888 nodes and 23,184 cores, is one of the most productive, multi-use cyberinfrastructure HPC and remote visualization resources in the world. Each compute node in this Dell/Linux cluster has 24GB of memory, and the login/development nodes have 16GB. System storage includes a 1,000-TB parallel (SCRATCH) Lustre file system and 276TB of local compute-node disk space (146GB/node). Lonestar also provides access to five large memory (1TB) nodes, and eight nodes containing two NVIDIA GPUs, giving users access to high-throughput computing and remote visualization capabilities respectively. A QDR InfiniBand switch fabric interconnects the nodes.
with 1,888 nodes and 23,184 cores, is one of the most productive, multi-use cyberinfrastructure HPC and remote visualization resources in the world. Each compute node in this Dell/Linux cluster has 24GB of memory, and the login/development nodes have 16GB. System storage includes a 1,000-TB parallel (SCRATCH) Lustre file system and 276TB of local compute-node disk space (146GB/node). Lonestar also provides access to five large memory (1TB) nodes, and eight nodes containing two NVIDIA GPUs, giving users access to high-throughput computing and remote visualization capabilities respectively. A QDR InfiniBand switch fabric interconnects the nodes.

Sacks and Hsu use two TACC supercomputers in their heart valve research, Stampede and Lonestar, both based on Intel technologies. Stampede is a Dell PowerEdge cluster equipped with Intel Xeon Phi coprocessors, 6,400 nodes, and 522,000 processor cores. Lonestar is a Dell/Linux cluster containing 1,888 nodes and 23,184 cores.

“On Stampede, our complex fluid-structure interaction simulations take three to five days to simulate, but they would take about 1.5 to 3 years to run on a single-core desktop system,” Hsu says. “That’s a speedup of about 200 times faster.”

Supercomputers deliver another kind of speedup—the ability to run multiple simulations in parallel. The team uses 256 cores for each simulation, which is not that much in the context of a supercomputer that contains hundreds of thousands of cores. But the ability to run 10 to 20 test cases simultaneously and get results in a few days enabled the ability to identify problems that led to fine-tuning the team’s mathematical formulas.

“If we were using a smaller-scale cluster, we’d need months to collect all these data,” Hsu says. “With the help of HPC, it became days. So even though one simulation requires only 256 cores, our study was only possible because we had access to more than 5,000 cores.”

“Our long-term goal is that our research will enable valve designers and surgeons to develop the next generation of durable heart valves,” Sacks adds. “Today, most heart valves are still developed based on years of empirical experience and have not reached optimal performance.  In addition, we have yet to exploit the possibilities of patient-specific designs.  The ability to incorporate actual predictive material modeling within the context of a fully functional heart valve simulation has never been achieved. When active, it will be a paradigm shift for the entire field.”

Hsu adds, “My research in computational mechanics takes me into several areas including wind energy and composites. However, in those fields we may be able to wait for answers. In the medical field, patients cannot wait. That’s why supercomputers are so critical to this field. Getting answers faster literally means the difference between life and death for some patients.”

Sacks’ and Hsu’s research team plans to investigate the use of the Intel Xeon Phi processors in Stampede to speed the linear solver part of the simulations, which is the most time-consuming part of the code. Another possible use for the coprocessor involves the team’s newer FSI code, which makes extensive use of third-party linear-solver libraries such as the Portable, Extensible Toolkit for Scientific Computation (PETSc), maintained by the Argonne National Laboratory.

TACC provides builds of common scientific libraries that are optimized for its hardware, and some library routines may be able to run computations on the Xeon Phi coprocessors. TACC also provides a support services team to help researchers parallelize and test code.

Sign up to our Newsletter

Featuring highlights, analysis, and stories from the week directly from us to your inbox with nothing in between.
Subscribe now

Be the first to comment

Leave a Reply

Your email address will not be published.


*


This site uses Akismet to reduce spam. Learn how your comment data is processed.