A comparison between plausible models in layered turbid media with geometrical variations applying a Bayesian selection criterion

Abstract

One possible application of Near Infrared techniques is to analyze human brain metabolic activity. Currently used models take into account the layered structure of the human head but, usually, they do not consider the non-planar surface of some of the boundaries, i.e. gray matter, which results in a much more complex structure, thus leading to more sophisticated models and longer calculation times. The main objective of this work is to determine if it is worth to replace a planar layered structure by a non-planar one. To this end we implement a Bayesian-based quantitative methodology for choosing between two competitive models describing light propagation in layered turbid media. Experiments of time-resolved diffuse reflectance measurements are performed in layered phantoms and complemented with numerical calculations. The resulting Distributions of Time of Flight of both models are compared using Bayesian model selection analysis. The non-planar interface was introduced in the simulations by a simple surface parametrization. Results suggest that, under certain conditions, a multilayer model with planar boundaries is good enough.