Simulating fluid flow within coronary arteries using parallelized sparse lattice bolztmann method

In clinical diagnostics of the coronary artery stenosis, the state of fluid flow is analyzed by
invasively measuring fractional flow reserve (FFR). However, a more detailed hemodynamic
evaluation of the flow and its spatial and temporal distribution may be more insightful for the
diagnostics. A realistic simulation of blood flow inside the patient-specific coronary artery could
be a better alternative to the invasive measurement of FFR. In this study, lattice Boltzmann (LB)
method was used for the simulations of fluid flow through complex geometries of human
coronary arteries. For such complex geometries, a large number of nodes in the LB mesh are
denoted as fixed and this imposes large computational and memory requirements that are
unnecessary. In this study, this standard approach is improved by using a sparse domain that
includes the fluid nodes and its closest fixed (solid) nodes. This way, the amount of required
memory resources is significantly reduced. During implementation, the software is parallelized to
execute on a GPU (graphics processing unit) device, ensuring that the amount of computational
time needed for the execution of the simulation is also significantly reduced. The software
presented in this paper provides a valuable tool that can be used in clinical practice for quick
assessment of the state of the arteries and provide fast quantitative information about fluid flow
that could be very useful for medical diagnostics and decision-making.