R. Montes Diez, A. Quirós
During the last few decades our knowledge of the human brain has developed significantly as a result of new neuroimaging techniques, such as functional magnetic resonance imaging (fMRI). By observing the relation between a stimulus paradigm and the changes in blood oxygenation in the brain, the so-called hemodynamic response function (HRF), fMRI provides a measure of brain activation. Modeling the HRF in fMRI experiments is therefore an important aspect of the analysis of data in functional neuroimaging. This has been done in the past using parametric response functions, typically Poisson, gamma or Gaussian densities. In this work, we consider the case in which the HRF is simply defined by an unknown function z(.). General Gaussian Processes theory presents an attractive way of expressing prior beliefs about the function z(.) and we show how, in this context, a combination of analytical methods may be used for making inference about the posterior predictive distribution of interest.
Palabras clave: fMRI, Gaussian processes
Programado
XC1a Pósters (Estadística)
18 de abril de 2012 12:00
Salón Madrid
