Duke’s Amanda Randles says her supercomputer simulations could help device makers

Amanda Randles, assistant professor of biomedical sciences at Duke University, has developed computational methods that enable the creation of ultra-realistic, 3D simulations of biological processes occurring in the human body, down to the cellular level.

Her work earned her the $250,000 Association for Computing Machinery (ACM) Prize in Computing (funded by Infosys), which honors early-stage and intermediate-level computer scientists who contribute to innovative research with fundamental impact and broad implications.

“Her early work included creating accurate 3D simulations of blood flow through the circulatory system,” ACM said of Randles, who is now the Alfred Winborne and Victoria Stover Mordecai Associate Professor of Biomedical Sciences at Duke University’s Pratt School of Engineering. “More recently, she and her team have developed biomedical simulations that have direct and tangible impact on patient care, including simulations of 700,000 heartbeats (the previous state of the art was 30 heartbeats), the interaction of millions of cells, and cancer cells moving through the body.”

The heartbeat simulation algorithm was developed using portable devices that collect data showing the state of a person’s circulatory system during normal activity, which is an improvement over the current method of taking photographs taken, for example, in a doctor’s office.

“Although still early in her career, Randles has led her field in developing computational tools that enable high-fidelity 3D blood flow simulations to diagnose and treat a variety of human diseases,” ACM said. “Her major achievements in this field include developing the first cellular simulation of the coronary artery tree for an entire heartbeat, using 1.5 million computer processing units (CPUs) to simulate blood flow on a whole-body scale, and using trained machine learning models to develop a framework for predicting key hemodynamic metrics under novel conditions.”

“She also developed a new way of modeling the human heart that allowed simulations of the heart for a large group of patients,” ACM continued. “In turn, these simulations led to a series of papers in which she showed that in order to model complex flow phenomena, it is necessary to consider the entire arterial tree, including its side branches. Randles’ full 3D simulations can also be used by cardiologists to plan therapeutic procedures. For example, these simulations can help physicians determine noninvasively which coronary artery lesions require treatment or, perhaps, how coronary artery hemodynamics might be affected by placing a rigid metal stent in a flexible artery.”

Medical Design and Outsourcing asked Randles to explain how her research can help device developers and manufacturers build and test innovative new products. The following text has been lightly edited for clarity and length.

MDO: Did you learn anything about these parts of the anatomy or how to develop these models/simulations that would be uniquely valuable to device developers?

Randles: “These types of models can be incredibly useful for device developers. We can simulate the deployment of devices of different shapes, structures, or sizes and assess the impact of design changes on the response of the individual. As a simple example, we previously assessed whether knowledge of a patient’s blood flow patterns would affect the physician’s choice of stent length. We showed that observing wall shear stress patterns did indeed statistically change the stent length recommended by the interventional cardiologists in the study. Additionally, by allowing the virtual surgeon to see the results of different device design decisions, we can provide a feedback loop for refinement or guidance, such as patients with specific anatomical features who may be better suited for a particular device or design. First, we really showed that vascular anatomy varies significantly across patients, and these differences have a significant impact on blood flow patterns and therefore on disease localization and progression. Quantifying these differences and taking them into account can be critical for device developers.”

MDO: Does this also apply to testing digital prototypes of cardiology, oncology or other devices?

Randles: “Yes, absolutely. These types of models provide a framework for testing different devices for a given patient before they even enter the operating room. With the new Longitudinal Hemodynamics Mapping Framework, you can even prospectively model the forces that the device will be subjected to over a period of months and determine different levels of activity.”

MDO: Can these models be personalized to a specific patient using data from wearable devices?

Randles: “Yes, personalization is a key part of everything we’re working on. Right now, we’re focusing on patient groups where we have medical images to provide 3D vascular anatomy of the patient, and then we’re doing a series of ongoing studies on the role of wearable data in driving these simulations over time. We’re using these models to create personalized digital twins of patients who have had carotid artery repair and congestive heart failure, for example.”

MDO: What are the biggest opportunities for device makers and manufacturers?

Randles: “The ability to virtually place a device in a patient’s arteries and see what forces it will be exposed to over periods of months or perhaps even years could provide valuable information to device developers. They can now assess how implementing the device will affect flow in the near future, but also test hypothetical scenarios, such as varying levels of sedentary lifestyles or potentially advanced comorbidities.”