Multilevel statistical inference from functional near infrared spectroscopy signals

Functional near infrared spectroscopy (fNIRS) is a technique that tries to detect cognitive activity by measuring changes in the concentrations of the oxygenated and deoxygenated hemoglobin in the brain. We develop Bayesian statistical tools for making multilevel inferences, that is, inferences generalizable to a population within the context of fNIRS neuroi-maging problem. Specifically, we present a method for multilevel modeling of fNIRS signals using a hierarchical general linear model. The model is treated in the context of Bayesian networks. Experimental results of a cognitive task (Stroop test) are presented with comparison to classical approaches.