A Bayesian signal detection procedure for scale-space random fieids

The authors consider the problem of searching for activation in brain images obtained from functional magnetic resonance imaging and the corresponding functional signal detection problem. They develop a Bayesian procedure to detect signals existing within noisy images when the image is modeled as a scale space random field. Their procedure is based on the Radon-Nikodym derivative, which is used as the Bayes factor for assessing the point null hypothesis of no signal. They apply their method to data from the Montreal Neurological Institute.