Detection of brain activation from magnitude fMRI data using a Generalized Likelihood Ratio Test

Functional magnetic resonance imaging (fMRI) measures the hemodynamic response in the brain that signals neural activity. The purpose is to detect those regions in the brain that show significant neural activity upon stimulus presentation. Most statistical fMRI tests used for this purpose rely on the assumption that the noise disturbing the data is Gaussian distributed. However, the majority of fMRI studies employ magnitude image reconstructions that are known to be Rician distributed, and hence corrupted by non-Gaussian distributed noise. In this work, we propose a Generalized Likelihood Ratio Test (GLRT) for magnitude MRI data that exploits the knowledge of the Rician distribution. The performance of the proposed GLRT is evaluated by means of Monte Carlo simulations.