Functional neuroimaging is a fast-developing area of research aimed at investigating the neuronal activity of the brain in vivo. The data for these studies are provided by positron emission tomography (PET) and functional magnetic resonance imaging (fMRI). Unfortunately, the analysis of these images is complicated by the variability of the data, by low signal-to-noise ratios, and by limited spatial resolution. Another major difficulty is that the shape and size of the brain varies from one individual to another. The purpose of this project is to develop image-processing algorithms for the quantitative analysis of such data sets. We have designed a general multiresolution procedure for the efficient registration of PET or fMRI volumetric data sets. Our geometrical transformation model allows for translation and general affine transformations (2-D or 3-D), including scaling and rotation. In order to improve the detection of significant differences between subjects, we have developed a method that performs statistical testing in the wavelet domain. The main advantage is that the discriminative information, which is smooth and well localized spatially, becomes concentrated into a relatively small number of coefficients. We have also designed new segmentation procedures to automate the stripping of the brain in MRI, and to detect the corpus callosum. These methods use a priori information on the shape of the objects; this knowledge is typically expressed in the form of connectivity constraints.
