B0280
Title: Bayesian semi-parametric modeling of near infra-red spectroscopy data
Authors: Timothy Johnson - University of Michigan (United States) [presenting]
Abstract: Functional near-infrared spectroscopy (fNIRS) is a relatively new neuroimaging technique. It is a low cost, portable, non-invasive method to monitor brain activity. Similar to fMRI, it measures changes in the level of blood oxygen in the brain. Its time resolution is much finer than fMRI, however its spatial resolution is much courser-similar to EEG or MEG. fNIRS is finding widespread use on young children whom cannot remain still in the MRI magnet and it can be used in situations where fMRI is contraindicated-such as with patients whom have cochlear implants. Since it measures the hemodynamic response to stimulus, similar to fMRI, statistical methods that are in use simply use modifications to existing fMRI packages. We show that these methods are inadequate and we propose a fully Bayesian semi-parametric hierarchical model to analyze fNIRS data. The hemodynamic response function is modeled using a cubic B-spline basis while nuisance signals (e.g. vasomotor signal and heart beat) are modeled using a Gaussian process. We assume the residual time-series is a high-order AR process and adopt a spike-and-slab prior to shrink unnecessary AR parameters to zero. Our model is easily adapted to handle the bivariate fNIRS time-series data at a single detector (oxygenated-and deoxygenated-hemoglobin). It can also easily be adapted to handle the spatial aspects of an array of detector.
Journal