Computational feasibility of simulating whole-organ vascular networks
Document Type
Article
Publication Date
10-21-2020
Abstract
The human body contains approximately 20 billion blood vessels, which transport nutrients, oxygen, immune cells, and signals throughout the body. The brain's vasculature includes up to 9 billion of these vessels to support cognition, motor processes, and myriad other vital functions. To model blood flowing through a vasculature, a geometric description of the vessels is required. Previously reported attempts to model vascular geometries have produced highly-detailed models. These models, however, are limited to a small fraction of the human brain, and little was known about the feasibility of computationally modeling whole-organ-sized networks. We implemented a fractal-based algorithm to construct a vasculature the size of the human brain and evaluated the algorithm's speed and memory requirements. Using high-performance computing systems, the algorithm constructed a vasculature comprising 17 billion vessels in 1960 core-hours, or 49 minutes of wall-clock time, and required less than 32 GB of memory per node. We demonstrated strong scalability that was limited mainly by input/output operations. The results of this study demonstrated, for the first time, that it is feasible to computationally model the vasculature of the whole human brain. These findings provide key insights into the computational aspects of modeling whole-organ vasculature.
Publication Source (Journal or Book title)
Biomedical physics & engineering express
First Page
055028
Recommended Citation
Donahue, W., & Newhauser, W. (2020). Computational feasibility of simulating whole-organ vascular networks. Biomedical physics & engineering express, 6 (5), 055028. https://doi.org/10.1088/2057-1976/abaf5b