MCDataset: a public reference dataset of Monte Carlo simulated quantities for multilayered and voxelated tissues...

Area of Science:

• Biomedical Optics
• Computational Physics
• Medical Imaging

Background:

• Open-source Monte Carlo (MC) methods for light propagation modeling are often complex to implement and require licensed software, hindering research and cross-validation in biomedical optics.
• Lack of standardized datasets limits the rigorous comparison and validation of different MC implementations.

Purpose of the Study:

• To propose an open-source tool (PyXOpto) for light propagation modeling.
• To create an accessible dataset for MC method comparison.
• To foster communication and consistency among researchers in biomedical optics.

Main Methods:

• Developed PyXOpto, an open-source MC implementation using Python and PyOpenCL for massively parallel computation on OpenCL-enabled devices.
• Modeled light propagation in multilayered and voxelated tissues.
• Generated a comprehensive dataset including reflectance, transmittance, energy deposition, and sampling volume for diverse configurations.

Main Results:

• PyXOpto demonstrated good agreement with a reference MC implementation.
• Identified a bias in the reference MC implementation attributed to its random number generator.
• The generated dataset facilitates validation and comparison of MC codes.

Conclusions:

• A common dataset is crucial for validating existing and developing new MC codes for light propagation in turbid media.
• PyXOpto offers a validated, open-source solution for light propagation modeling.
• Addressing biases in random number generators is important for MC accuracy.