Given a collection of images describing the same objects, one should
be able to build a good model autonomously. Given sets of images from
normals and patients with a common pathology, different atlases will
be constructed and diagnosis become solvable by computers.