Presentation Information
[C18-03]Multiscale Modeling of an Entire Human Blood Circulation Network with Parallel Computing
*Jiawei Liu1, Hiroshi Suito1 (1. Tohoku University (Japan))
Keywords:
blood circulation network,multiscale model,parallel computation,cerebrovascular bypass simulation
We present a multiscale model to simulate the dynamics of blood circulation within a global closed-loop network encompassing arterial, venous, and portal venous systems, along with heart-pulmonary circulation and capillary micro-circulations. One-dimensional (1D) models simulate large blood vessel flow, whereas zero-dimensional (0D) models are used for vascular subsystems corresponding to peripheral arteries and organs (e.g., liver, stomach, spleen, pancreas, intestine), which play important roles in metabolic system dynamics. This integrated 1D-0D multiscale model captures hemodynamic characteristics across both spatial and temporal domains.
A key innovation of this work is the implementation of an iterative procedure that allows for parallel computations, prioritizing both execution time and computational stability through the incorporation of shortest-job-first (SJF) scheduling and interval partitioning strategies. Our analysis indicates that the proposed efficient parallel algorithms for multicore environments solve the equation systems much faster than serial computations. Global multiscale models of the entire blood circulation networks, which encompass numerous blood vessels and junctions, typically demand substantial computational resources. Therefore, our emphasis on high-efficiency computations for this constructed multiscale model is crucial for achieving real-time simulations.
Furthermore, we apply the model to quantitatively evaluate hemodynamic changes in cerebrovascular networks before and after bypass surgery, using medical imaging data. By integrating the multiscale model with a depth-first-search (DFS) algorithm, we accurately simulate flow dynamics under varying cerebrovascular structures and inflow conditions. Moreover, simulation results confirm the presence of to-and-fro flow in certain vessels, highlighting the influence of cerebral architecture and perfusion dynamics on blood circulation. This framework offers a robust tool for early diagnosis and surgical planning in clinical settings.
A key innovation of this work is the implementation of an iterative procedure that allows for parallel computations, prioritizing both execution time and computational stability through the incorporation of shortest-job-first (SJF) scheduling and interval partitioning strategies. Our analysis indicates that the proposed efficient parallel algorithms for multicore environments solve the equation systems much faster than serial computations. Global multiscale models of the entire blood circulation networks, which encompass numerous blood vessels and junctions, typically demand substantial computational resources. Therefore, our emphasis on high-efficiency computations for this constructed multiscale model is crucial for achieving real-time simulations.
Furthermore, we apply the model to quantitatively evaluate hemodynamic changes in cerebrovascular networks before and after bypass surgery, using medical imaging data. By integrating the multiscale model with a depth-first-search (DFS) algorithm, we accurately simulate flow dynamics under varying cerebrovascular structures and inflow conditions. Moreover, simulation results confirm the presence of to-and-fro flow in certain vessels, highlighting the influence of cerebral architecture and perfusion dynamics on blood circulation. This framework offers a robust tool for early diagnosis and surgical planning in clinical settings.