Accelerating mesh-based Monte Carlo photon transport simulation using ray-tracing hardware (Conference Presentation)

Mesh-based Monte Carlo (MMC) methods have been broadly adopted in biophotonics as a quantitatively accurate light modeling technique in complex media, including biological tissues. By utilizing triangular and tetrahedral meshes for domain representations, MMC methods overcome the limitations of traditional layered or voxel-based MC methods in handling curved tissue boundaries, yielding substantially improved accuracy. Over the last few years, massively parallel MMC algorithms, powered by graphics processing units (GPUs), have been reported to offer both high simulation speed and accuracy. In this work, we explore the real-time ray-tracing (RT) hardware in contemporary GPUs and report a further accelerated RT-MMC algorithm that achieves 1.5x to 4.5x speedups over conventional software-based GPU-accelerated MMC. The RT-MMC algorithm is implemented using the NVIDIA OptiX ray-tracing library, extending the ray-tracing workflow originally aimed for computer graphics to 3-D quantitative photon transport modeling. We have validated this RT-MMC method in both simple and sophisticated domains and characterized the speedup across various GPU hardware architectures.