An Ontological and Non-monotonic Rule-based Approach to Label Medical Images

Medical images can nowadays be automatically segmented but a semantic identification of their parts remains an open question. We think an approach based on the integration of OWL ontologies and SWRL rules can be applied to model medical knowledge and to label medical images. Nevertheless, negation and non-monotonic operators are not included in SWRL language and so we can not express modeling exceptions, cope with a dynamically changing knowledge and make the closed-world assumption. As a result, we have extended SWWL language to express i) a weak form of Negation as Failure and ii) a non-monotonic operator for statement removal. Besides, we have realized an ontology-based service that i) exploits and integrates ontologies and rules in a homogenous system, ii) performs both monotonic and nonmonotonic reasoning. Finally, as an application, we have used the Ontology Service to label the brain anatomical structures and to recognize the brain abnormalities due to polymicrogyria.