Multi-center neuroimaging studies have more power than smaller ones to conduct sophisticated studies of basic neuroanatomy and clinical disorders. One important confound of combining images obtained from different scanners is the potential for scanner effects to introduce systematic error, thus making the interpretation of results difficult. Differences in diffusion imaging measurements due to scanner-dependent inaccuracies may either mimic or obscure true changes. In the context of multicentric Swiss project on brain connectivity in patients with epilepsy, we used DSI (Diffusion Spectrum Imaging) acquired on different scanners to address the limitation of DTI (Diffusion Tensor Imaging) where imaging of multiple fiber orientation in a single voxel is not possible. Since DSI is a very recent development of MRI there is no information on inter- and intra-site reproducibility, while concerning the DTI there are very few studies on a 3T scanner and even less on cross center reliability measures.