Amanda Randles

Amanda Randles is at the forefront of creating vascular digital twins, bringing computational health advancements to the forefront of medical practice. While cardiovascular research has grown, the medical community still faces challenges in developing precise and detailed models of vascular systems for clinical application. Randles' work addresses these gaps by pushing the boundaries of what can be achieved with high performance computing (HPC) and cloud technology in simulating 3D hemodynamics. Her contributions not only enhance our understanding of vascular dynamics but also strive to expand the spatial and temporal scales at which these systems can be modeled, critical steps towards their clinical translation. Her work provides a strong foundation for using computational tools to aid in diagnosing and treating human disease.

Expanding the scope of domain modeled: Her first major achievement in computational blood flow models was in 2010 when her team conducted the inaugural simulation of the coronary arterial tree at a cellular resolution across a full heartbeat.  She built on this experience by developing HARVEY, a massively parallel computational fluid dynamics software for calculating patient-specific blood flow maps. The code was named after William Harvey, who first discovered the circulation of blood. By utilizing 1.5 million computing cores of the Lawrence Livermore National Laboratory's Sequoia supercomputer, she performed the first full-body scale simulation of 3D blood flow. Randles' leadership in method development has further optimized the use of heterogeneous supercomputers and cloud computing resources, culminating in multiscale models significantly reducing computational demands. The resulting reductions in both time and energy make clinical translation of such tools easier.

In her work with HARVEY, Randles and her team introduced a novel moment-based representation technique that reduces memory usage, thus reducing the computational hardware required for high-resolution simulations.  This innovation allows for a larger patient cohort to be modeled.  Furthermore, her research has resulted in a suite of influential publications underscoring the necessity of detailed models encompassing the complete 3D arterial tree including side branches to accurately reflect the nuances of blood flow and heart disease. These enhanced models offer more precise, personalized simulations and can enable clinical trials to cover conditions previously excluded.  Randles' work has helped quantify when complete 3D models are needed as opposed to reduced order alternatives.  

Bridging scales through multiscale modeling: Randles has significantly contributed to fluid-structure-interaction modeling as well. Randles and her team crafted a computational method tailored to distinct cell types, confirming the model's accuracy against microfluidic experimental data for both cancerous and red blood cells. Her team then pioneered techniques that scale to simulate hundreds of millions of cells using leadership class heterogeneous computing architectures.  Tackling the computational challenge of simulating cell behavior across vast spatial scales, from the microscopic to the potentially macroscopic, her team has developed the Adaptive Physics Refinement (APR) method.  This innovative approach allows for the study of cellular interactions over large distances that could determine why certain cancer cells metastasize to specific locations. Traditional models, even when run on the most advanced supercomputers, are limited to representing a few cubic millimeters with detailed 3D deformable cells.  The APR method surmounts this limitation without compromising accuracy. This advancement is important for elucidating underlying mechanisms driving disease and could significantly impact our understanding of disease initiation and progression.

Randles' further contributions to the field of computational health include methods integrating machine learning and physics-based simulations. She specifically addresses how comorbidities like anemia or hypertension impact hemodynamics in cardiovascular diseases. Her work using massively parallel computing offers predictive insights into the exacerbating effects of exertion and other physiological factors on patients with conditions such as coarctation of the aorta.

Building upon these methodologies, Randles' team created novel longitudinal hemodynamic maps by developing an algorithm that extends the time frame of simulations from seconds to weeks. Introducing a novel computational method to drive the flow simulations from wearable biosensor data, this work lays the foundation for a comprehensive view of a patient's vascular health over time, driven by wearable biosensor data, thus pioneering a new direction in computational health.

Amanda has established an ambitious research program focused around the transformation of her circulatory modeling code HARVEY into a world-class research tool capable of making a direct clinical impact. HARVEY is a massively parallel circulatory simulation code capable of modeling the full human arterial system at subcellular resolution. Her work leverages high performance computing to solve biomedical problems. She has developed large-scale computational methods to study disease localization and progression to foster discoveries that will serve as a basis for improving the diagnosis, prevention, and treatment of human diseases.

Together with an interdisciplinary team of doctors, scientists, and computer scientists Amanda has developed several new methods of high-performance computer simulations for blood flow in arteries across multiple scales  from molecules to blood cells. She has been applying theories from cosmology to the simulation of the forces from the beating heart. She developed her own parallel application code (HARVEY) to model flow in microfluidic devices and aneurysms as well as in coronary arteries. Amanda's multi-disciplinary background in applied physics and of computational methods and parallelization strategies in computer science, serves her well in these endeavors.

Amanda has remained keenly aware of the need to translate computational results into actionable data doctors can use to improve patient outcomes, and has continued to build collaborations that will allow her use this new computational capability to bridge the gap from the computer to the clinic. Her cross-disciplinary approach will yield new insights into efficient fluid flow simulations for complex multi-scale systems that will be useful well beyond blood flow simulations.