GPU-based Monte Carlo simulation for light propagation in complex heterogeneous tissues.

Area of Science:

• Biomedical Optics
• Computational Imaging
• Medical Physics

Background:

• Monte Carlo (MC) simulations are crucial for modeling light propagation in biological tissues, particularly in optical molecular imaging.
• The computational expense of traditional MC methods limits their application due to the complexity of heterogeneous tissues and the large number of photon simulations required.

Purpose of the Study:

• To develop and evaluate a parallelized MC simulation for light propagation in heterogeneous tissues.
• To accelerate MC simulations on graphics processing units (GPUs) for improved efficiency in optical molecular imaging.

Main Methods:

• A parallel MC simulation code was implemented using the Compute Unified Device Architecture (CUDA) platform on GPUs.
• Optimizations were employed to minimize memory access latency by utilizing GPU constant and texture memory.
• The implementation was tested on homogeneous and heterogeneous mouse models using an NVIDIA GTX 260 GPU and an Intel Xeon CPU.

Main Results:

• The parallel MC simulation on GPU demonstrated feasibility and significant efficiency gains compared to traditional methods.
• The optimized code effectively handles complex tissue geometries defined by triangle meshes.

Conclusions:

• Parallel MC simulation on GPUs offers a viable and efficient approach for modeling light propagation in heterogeneous tissues.
• This advancement can accelerate research and applications in optical molecular imaging and related fields.