Computational models are essential for assessing quantities that are otherwise immeasurable. In general, our work focuses on the design of large-scale parallel applications targeting problems in physics. We design large-scale parallel applications that enable the study of research problems in areas ranging from cardiovascular disease to drug development. One of main goals is to develop and deploy a general purpose, multiscale methodology to study blood flow patterns in complex environments relating to real patient arterial geometries as measured by CT or MRI imaging.
The recognition of the role hemodynamic forces have in the localization and development of disease has motivated large-scale efforts to enable patient-specific simulations. When combined with computational approaches that can extend the models to include physiologically accurate hematocrit levels in large regions of the circulatory system, these image-based models yield insight into the underlying mechanisms driving disease progression and inform surgical planning or the design of next generation drug delivery systems. The scale of these simulations requires the use of massively parallel supercomputers, so much of our work involves the development of methods to maximize parallel efficiency. The scope of projects includes only vascular diseases, but also treatment planning and the movement of circulating tumor cells in the bloodstream. Predicting the location of secondary tumor sites is a critical hurdle in the understanding and treatment of cancer. The goal of this research is to develop a method of predicting likely sites of cancer metastasis using a combination of personalized massively parallel computational models and experimental approaches.
We are building massively parallel patient-specific simulations to study the development and treatment of disease. The image below shows the domain decomposition of a section of a patient's aorta. Each square represents the bounding box handled by a different processor. The image is courtesy of Liam Krauss, LLNL.