Amanda Randles: Computing Complex Biological Systems

Supercomputers model blood flow to improve care outcomes

Amanda Randles, a pioneer in using supercomputers to gain insight into biomedical challenges, will join Duke University’s Biomedical Engineering Department on July 1, 2015. Bringing with her computer models of how blood, particles and cells travel through our veins and arteries, Randles will work with Duke Medicine physicians to identify the best ways to treat patients with ailments like aneurysms, cancer and coronary artery disease on a case-by-case basis. 

This line of research fuses two of Pratt’s burgeoning strengths together—the strong ties between engineering and Duke Medicine and the newly endowed Information Initiative at Duke (iiD)—and opens a new line of research that many are sure to follow.

“The iiD is delighted to have helped the Biomedical Engineering Department bring Amanda to Duke, and we are looking forward to developing new partnerships with Pratt,” said Robert Calderbank, the Charles S. Sydnor Professor of Computer Science, a professor of electrical and computer engineering and the director of the iiD. “Her expertise in large-scale computing brings a new capability to the iiD community.”

“Amanda will help Duke’s Biomedical Engineering Department be one of the first to take the plunge into high performance computational bioengineering,” echoed Ashutosh Chilkoti, director of Duke’s Biomedical Engineering Department. “We are very excited to help chart this new and exciting course for the field.”

Randles’s appointment is a coming home of sorts, as she graduated from Duke University with dual bachelor’s degrees in computer science and physics in 2005. While focusing on computational methods in genetics and bioinformatics, she wrote her undergraduate thesis on optics under the tutelage of David Brady, professor of electrical and computer engineering, and Bob Guenther, adjunct professor of physics.

After three years at IBM as a developer on its Blue Gene project—the world’s fastest supercomputer at the time—Randles returned to school so she could conduct more research on the supercomputers she was helping to develop. She enrolled at Harvard University, where she earned a master’s degree in computer science and a PhD in applied physics. After moving to Lawrence Livermore National Laboratory to work with some of the country's largest supercomputers, Randles was able to fully fund her cardiovascular modeling project last year and found herself ready to start her own research enterprise.

She chose to return to Duke among many enticing offers—and not just for the basketball (although she’s known to be a diehard Blue Devil).

“I’ve spent the past few years focusing on the computer science required to develop a 3D model of the cardiovascular system, and now I want to focus on using my work to make a medical impact,” said Randles. “Being able to walk to a hospital and meeting physicians who explicitly want to work with engineers is exactly what I was looking for, and that’s what I found at Duke.”

Randles specializes in writing code for supercomputers with thousands of processors that work in parallel. By splitting the computational tasks among them simultaneously, researchers can cut calculations that would otherwise take weeks down to days or even hours.

The code that Randles is cultivating creates 3D models of a patient’s specific blood-flow dynamics. With this information in hand, doctors can noninvasively measure risk factors and choose the best surgical or interventional option for that patient. For example, the model could indicate blood vessels with areas of low shear stress, which is an indicator of where plaque may build up along the wall of an artery.

The video below highlights some of the 3D models she developed while working on her PhD in applied physics in the research group of Efthimios Kaxiras, the John Hasbrouck Van Vleck Professor of Pure and Applied Physics at Harvard University.

Randles’s work can already create the calculations needed to model a small segment of the cardiovascular system, and she has been using 3D-printed models to test her predictions. After moving to Duke, however, she wants to begin working with physicians to verify that her program is accurate in actual biological systems.

“We’ve only recently gotten to the point where CT or MRI scans are detailed enough to even try to model blood flowing through the 3D geometry of the vasculature system,” said Randles. “And with supercomputers that were originally dedicated to physics or Department of Defense experiments now starting to be used by biomedical engineers, I think we’re only a few years away from being able to help real people.”

The second prong of Randles’s work, however, will take much longer to see through to fruition.  Once the small segments of her model have been verified, Randles wants to scale her work up to model the entire vascular system. From there, she hopes to use fluorescent metastatic cancer cells to determine what physical features of a person’s blood flow affect where a new tumor will start to grow.

“Realizing that dream is at least a decade or two away, though,” said Randles. “The supercomputers I will need to run those calculations haven’t even been built yet. But they’re coming, and I’ll be ready when they arrive.”