Efficient GPU-based Monte-Carlo simulation of diffusion in real astrocytes reconstructed from confocal microscopy

Area of Science:

• Computational biology
• Biophysics
• Medical imaging

Background:

• Understanding intracellular diffusion is crucial for cell biology.
• Current simulation methods struggle with complex cellular geometries.
• Confocal microscopy provides detailed cellular structure data.

Purpose of the Study:

• To develop an efficient Monte Carlo simulation for diffusion-weighted signals in complex cellular structures.
• To optimize particle trajectory simulations using novel computational approaches.
• To leverage GPU acceleration for enhanced modeling performance.

Main Methods:

• Utilized an octree structure for spatial decomposition of surface meshes.
• Implemented a radius-search algorithm to accelerate particle-face encounter identification.
• Developed a 3D binary marker for efficient cellular structure description and trajectory optimization.
• Created a GPU-based implementation of the octree and binary marker methods.

Main Results:

• The octree structure and radius-search algorithm significantly sped up Monte Carlo simulations.
• The 3D binary marker approach optimized particle trajectory simulations.
• The GPU-based binary marker method demonstrated unparalleled performance.
• The developed simulation accurately models diffusion in complex cellular environments.

Conclusions:

• The GPU-based binary marker approach offers a highly efficient method for intracellular diffusion modeling.
• This work enables better understanding of intracellular diffusion dynamics.
• The simulation can be used to validate diffusion models and generate diffusion signature dictionaries.