Our computational lab at Duke University focuses on using advanced computational methods to better understand and improve human health. A key aspect of this work involves designing large-scale parallel applications that allow us to study a variety of research problems, spanning numerical and computational to biomedical challenges. Our central aim is to develop and implement a multiscale methodology that can be used to analyze blood flow patterns in real patient arterial geometries, as captured by CT or MRI imaging.
One of the central components of this work is the creation of 3D realistic digital twins of the human vasculature. These digital twins provide us with a powerful tool for conducting simulations that accurately reflect real-world conditions and can be used to gain insights into the underlying mechanisms driving disease progression. By studying these digital twins, we hope to inform the design of effective treatments and therapies for a range of health problems. These simulations require the use of massively parallel supercomputers, which is why much of our work focuses on maximizing parallel efficiency.
We collaborate closely with clinicians and other researchers to develop accurate and robust computational models that are informed by clinical and experimental data. Our projects range from studying the diagnosis and treatment of vascular diseases to investigating the movement of circulating tumor cells in the bloodstream. The image below (created by Liam Krauss at LLNL) provides a glimpse into our work, showing the domain decomposition of a section of a patient's aorta, with each square representing the bounding box handled by a different processor.
