Modeling Nonlinear Transformations Which Reduce The Approximation Error Between Photon Transport Models in Diffuse Optical Tomography

Abstract

Diffuse Optical Tomography (DOT) is an emerging medical imaging technique. This method of imaging has incredible potential for global impact on access to healthcare due to low cost compared to other imaging modalities (X-Ray, CT, etc.). DOT is safer than the aforementioned modalities, and utilizes near-infrared light (NIRS), which is not harmful for humans. DOT is currently impractical because it produces scans that are not accurate enough for medical diagnoses. One point of inaccuracy is that NIRS is highly scattering while propagating through biomedical tissue (photons take a random variety of paths), so it is extremely difficult to find a model of this propagation that can support accurate image reconstruction. This project compares the Monte Carlo Model and Diffusion Equation Model of photon transport. The Monte Carlo Model (MCM) is well known to be extremely accurate, however is too computationally intensive to put into practice. The Diffusion Equation Model is a partial differential equation approximation of the MCM, so it is less accurate, however more efficient. By using the ValoMC software to simulate the MCM and the Toast++ software to simulate the Diffusion Equation Model, both with a variety of parameters, a pattern of approximation error in solutions of photon fluence can be found and modeled. The main objective is to use these models of approximation error to transform Toast++’s raw solutions to be more similar to the ValoMC solutions, and therefore more accurate.

GANIT J. Bangladesh Math. Soc. 45.2 (2025) 045–053