Coarctation of the aorta (CoA) or the narrowing of the main artery which brings blood from the heart to the rest of the body, is one of the most prevalence heart defects in the US, affecting over 1600 newborns each year. CoA can lead to many other health issues including premature coronary artery disease, stroke, hypertension, cardiac failure and aneurysm or the swelling of the artery wall, if it is not detected and treated early.

Early on, the former Lawrence Fellow and now Alfred Winborne and Victoria Stover Mordecai Assistant Professor of Biomedical Sciences at Duke University Amanda Randles and her mentor, computer scientist Erik Draeger at the Lawrence Livermore National Laboratory (LLNL) combined machine learning, 3-D printing and high performance computing simulations to precisely model blood flow in the aorta.

The largest ever CoA simulation study

Randles and her research team wished to understand the impacts of various physiological factors such as exercises, altitude changes and pregnancy, which demand the heart to pump harder to bring blood throughout the body, on CoA. The project was first proposed as an Institutional Computing Grand Challenge project at LLNL and it is the largest simulation of CoA to date involving more than 70 million compute hours of 3D simulations being run on the LLNL’s Blue Gene/Q Vulcan supercomputer.

The simulations relied on a fluid dynamics software called HARVEY, developed by Randles to mimic blood flow. HARVEY derived at a 3D geometries of the aorta using CT and MRI scans but because the aorta is large and complex, Randles, who has a background in high performance computing and biomedical simulation, decided to rewrite HARVEY for the supercomputer, so that the massive amount of simulations can be run accurately.

The research team began by running simulations of the aorta with stenosis, the narrowing of the left side of the heart which generates a pressure gradient in aorta and the rest of the body. They then studied the effects of different degree of stenosis, rate of blood flow and viscosity and use them to predict two diagnostic metrics: pressure gradient across the stenosis and stress experienced on the aorta wall, to reflect the kind of impact a patient’s lifestyles choices may have on CoA.

The role of machine learning and 3D printing

Researchers believe these simulations will facilitate a more realistic understanding of how CoA is affecting patients beyond examining them when they were sitting in the physician’s office or with just one simulation. Nevertheless, the impacts of various physiological factors are not linear. So, the number of simulations that needed to be run ought to be huge, especially the research team is interested to know how velocity and viscosity of blood flow at different points of the aorta.

Thus, to keep the number of simulations required to capture the impacts from all the physiological factors at a minimal, researchers decided to deploy machine learning. They trained the model using data gathered from all the 136 blood flow simulations performed on the supercomputer. The method allowed them to reduce the number of viscosity and velocity pairing from a few hundred to below ten. This will enable them to develop a more specific risk profile for each patient in the near future.

Furthermore, in order to validate the machine learning-driven simulation models, a separate group of researchers at the Arizona State University 3D printed aortas to complete benchtop experiments that can compare with the simulation results. In the long run, Randles’ research team hopes that a synergy between machine learning, high performance computing and 3D printing validation design will shed new lights on certain risk factors that are worth prolong monitoring.

All study results were published in journal Scientific Reports. If you are interested in how technology, particularly, artificial intelligence (AI), robotics, virtual/augmented/mixed realities, and many others are influencing Cardiology and strategies to deploy them in the clinical setting, do not miss the upcoming AIMed Cardiology virtual event, organized in association with the American College of Cardiology, taking place on 4 November. Register your interest or get a copy of the agenda here today!


Author Bio

Hazel Tang A science writer with data background and an interest in the current affair, culture, and arts; a no-med from an (almost) all-med family. Follow on Twitter.