Accelerated rescaling of single Monte Carlo simulation runs with the Graphics Processing Unit (GPU)

Area of Science:

• Biomedical Optics
• Computational Biology
• Medical Imaging

Background:

• Accurate interpretation of light remitted from biological tissues is crucial for clinical applications.
• Traditional analytical and stochastic models (e.g., Monte Carlo) for light transport are computationally intensive.
• Real-time measurement of tissue optical properties necessitates accelerated simulation methods.

Purpose of the Study:

• To develop a Graphics Processing Unit (GPU)-based method for accelerating Monte Carlo simulations.
• To enable rapid calculation of diffuse reflectance values for various tissue optical properties.
• To provide an accessible, open-source software package for researchers.

Main Methods:

• Utilized Graphics Processing Units (GPUs) for accelerating rescaling of single Monte Carlo runs.
• Developed a four-abstraction-layer software package in MATLAB for ease of use by non-specialists.
• Implemented a GPU-based approach to rapidly calculate diffuse reflectance for homogeneous tissues.

Main Results:

• Achieved a reduction in computation time of several orders of magnitude compared to other GPU-based methods.
• Generated diffuse reflectance values in as little as 0.08 milliseconds per simulation.
• Demonstrated processing speeds approximately 3400 times faster than alternative GPU approaches for multiple calculations.

Conclusions:

• The GPU-based rescaling approach significantly accelerates Monte Carlo simulations for diffuse reflectance.
• This method holds promise for real-time optical property measurements in clinical settings.
• Further optimization is needed to address CPU-to-GPU memory transfer time limitations.