Hierarchical mutual information clustering for an improved classification of fMRI data

Introduction: In functional magnetic resonance imaging (fMRI) the blood oxygenation level dependent (BOLD) effect is used to identify and delineate neuronal activity. Statistical analysis methods are commonly used to localize activated brain regions. For each voxel in the fMRI data set a statistical value is derived to build a statistical parametric map (SPM), which indicates possible neuronal activations. Nevertheless, these methods do not allow a finer classification of activated brain regions. Methods: In order to achieve a more detailed classification of fMRI data we estimate the mutual information (MI) cont…