To cope with the difficulty of 3D MRI brain scans segmentation, specification and instantiation of a priori models should be constrained by local images characteristics. We introduce situated cooperative agents for the extraction of domain and control knowledge from image grey levels. Their dedicated behaviors, i.e segmentation of one type of tissue, are dynamically adapted function of their position in the image, topographic relationships and radiometric information gradually gained during local region growing processes. Acquired knowledge is gathered and shared via qualitative maps. Incremental refinement of the segmentation is obtained through the combination, distribution and opposition of solutions concurrently proposed by the agents.
展开▼