Digital Blood in Massively Parallel CPU/GPU Systems for the Study of Platelet Transport (1911.03062v1)

Abstract: We propose a highly versatile computational framework for the simulation of cellular blood flow focusing on extreme performance without compromising accuracy or complexity. The tool couples the lattice Boltzmann solver Palabos for the simulation of the blood plasma, a novel finite element method (FEM) solver for the resolution of the deformable blood cells, and an immersed boundary method for the coupling of the two phases. The design of the tool supports hybrid CPU-GPU executions (fluid, fluid-solid interaction on CPUs, the FEM solver on GPUs), and is non-intrusive, as each of the three components can be replaced in a modular way. The FEM-based kernel for solid dynamics outperforms other FEM solvers and its performance is comparable to the state-of-the-art mass-spring systems. We perform an exhaustive performance analysis on Piz Daint at the Swiss National Supercomputing Centre and provide case studies focused on platelet transport. The tests show that this versatile framework combines unprecedented accuracy with massive performance, rendering it suitable for the upcoming exascale architectures.

• The paper presents a hybrid CPU/GPU computational framework using LBM, FEM, and IBM to simulate cellular blood flow and platelet transport.
• The framework efficiently allocates tasks to CPU and GPU resources, showing strong performance and scalability on supercomputing platforms.
• This research provides detailed insights into platelet transport under varying red blood cell conditions, with implications for clinical applications and biomechanical modeling.

Analysis of Digital Blood for Parallel CPU/GPU in Platelet Transport Studies

The paper by Kotsalos et al. presents a computational framework designed for simulating cellular blood flow, with a particular focus on the transport mechanisms of platelets. This work introduces a versatile system offering enhanced performance while maintaining accuracy and complexity appropriate for the generation of numerical results applicable to biological phenomena studies.

In addressing the fluid dynamics of blood, the authors employ a hybrid methodological approach, integrating the lattice Boltzmann method (LBM) for plasma simulation, a finite element method (FEM) for modeling blood cell dynamics, and the immersed boundary method (IBM) for fluid-solid interaction. This approach effectively capitalizes on the strengths of each method to provide a robust simulation capable of mapping each phase of the transportation process.

Core Contributions

1. Hybrid Framework Design: The integration of CPU and GPU phases within the simulation environment allows for optimized performance across different computing platforms. The modular nature of the framework supports substitutability, thereby allowing adaptations to suit various computational needs and advancements.
2. Methodological Efficiency: The lattice Boltzmann solver Palabos is used for fluid simulation, while the npFEM solver is allocated to the GPUs for simulating deformable blood cells. Such an allocation enhances overall computational efficiency, notably improving performance metrics by maintaining low execution times for solid solver operations.
3. Performance Scalability: Through the incorporation of the Piz Daint supercomputer at the Swiss National Supercomputing Centre, the framework's performance was validated to be effective for both small-scale and mid-scale domain blood simulations. It showed promise for adaptation to exascale computing, suggesting that it can handle a significant increase in the problem space without proportional loss of performance.
4. Numerical and Biological Insights: The framework provided detailed insights into platelet transport across varying rheological conditions of red blood cells (RBCs). By simulating domains with variable hematocrit levels, the paper can project results applicable to both physiological and pathological blood conditions.

Implications and Future Directions

The implications of this research are multifaceted. Practically, this framework positions itself as a potent tool in the computational exploration of blood dynamics, critical for developing clinical applications like drug delivery systems and lab-on-chip devices. On a theoretical level, the paper advances the methodological discourse surrounding hybrid computational modeling in biomechanical contexts.

Moreover, the paper emphasizes future developments in automated high-performance computing (HPC) frameworks. The exploration of multi-scale modeling could potentiate layered simulations where first principles in high fidelity models guide more abstract simulations, broadening the practical applicability across physiological, medical, and bioengineering fields.

Conclusion

This paper puts forth a computationally efficient approach to decipher platelet dynamics using advanced hybrid CPU/GPU methodologies. It provides requisite numerical accuracy and detail necessary for next-generation research into cellular blood flow. As new computational paradigms emerge, further refinement and scaling of this framework could unlock deeper insights into the mechanical and pathological states of blood transport, contributing significantly to both theoretical development and practical innovation in biomedical research. The research sets the groundwork for a model-integrated platform possibly aligning with virtual physiological experimentation in the computation of human biology.