High level mammographic information fusion for real world ontology population

In this paper, we propose a novel approach for ontology instantiating from real data related to the mammographic domain. In our study, we are interested in handling two modalities of mammographic images: mammography and Breast MRI. Firstly, we propose to model both images content in ontological representations since ontologies allow the description of the objects from a common perspective. In order, to overcome the ambiguity problem of representation of image's entities, we propose to take advantage of the possibility theory applied to the ontological representation. Second, both local generated ontologies are merged in a unique formal representation with the use of two similarity measures: syntactic measure and possibilistic measure. The candidate instances are, finally, used for the global domain ontology populating in order to empower the mammographic knowledge base. The approach was validated on real world domain and the results were evaluated in terms of precision and recall by an expert. [ABSTRACT FROM AUTHOR]