Bayesian estimation of optical properties of the human head via 3D structural MRI

Knowledge of the baseline optical properties of the tissues of the human head is essential for absolute cerebral oximetry, and for quantitative studies of brain activation. In this work we numerically model the utility of signals from a small 6-optode time-resolved diffuse optical tomographic apparatus for inferring baseline scattering and absorption coefficients of the scalp, skull and brain, when complete geometric information is available from magnetic resonance imaging (MRI). We use an optical model where MRI-segmented tissues are assumed homogeneous. We introduce a noise model capturing both photon shot noise and forward model numerical accuracy, and use Bayesian inference to predict errorbars and correlations on the measurements. We also sample from the full posterior distribution using Markov chain Monte Carlo. We conclude that ∼ 10⁶ detected photons are sufficient to measure the brain's scattering and absorption to a few percent. We present preliminary results using a fast multi-layer slab model, comparing the case when layer thicknesses are known versus unknown.