New imaging modalities like MR diffusion imaging provide more and more clues about brain architecture.
This talk will discuss the idea that this new information will require the launching of a new research program
aiming at the computerized inference of new structural models of brain architecture.
These models are made up of a list of anatomical entities and their connectivity-based relationships.
The basic content of this program is illustrated first with a pure computer vision issue,
the inference of a structural model of chair. Then the nature of this program is detailed
using the old problem of modeling the cortical folding patterns. Finally, it is shown that various neuroimaging
issues fit within this framework: from connectivity-based parcellation of grey matter or fiber-based parcellation of white matter to structural group analysis of a set of individual activation maps.