Modeling, representing and sharing anatomical knowledge on brain cortex

In the context of medical knowledge modeling, this thesis proposes an ontology of the human adult cerebral cortex anatomy. This ontology has to be shareable and reusable in various application contexts, by humans as well as software.

The modeling step lays emphasis on rigorous modeling principles, on an explicit representation of the meaning and on semantic consistency. It is based on existing symbolic models of anatomy although they do not describe the brain cortex, and on theoretical works on knowledge modeling that are focused on taxonomy and mereotopology. Our approach relies on a strict distinction between the relationships representing specialisation, composition and spatial organization of anatomical structures. In addition to the previous works, the properties of each relationship are explicitely defined. We then propose an original method based on these properties for managing the semantic consistency of the model.

Reusing the model in various application contexts generates constraints on the computational properties of its representation. According to the needs of the model, we analyzed and compared the expressivity of several representation formalisms. The results highlight the necessity of a paradigm shift promoting the use of an highly expressive formalism during modeling, and the generation of various simplified representations matching the requirements of multiple applications.

Finally, Web technologies facilitate the reuse of the model in various application contexts. Specifically, Web services provide a standard framework to access the model concepts and relationships, as well as to share inference processing.

Keywords: ontology, anatomy, cerebral cortex, taxonomy, mereology, topology, semantic consistency, Internet, Web services.