Design and Modeling of Object Oriented Bayesian Network for Medical Applications.

Probabilistic graphical models (PGMs) are powerful tools for representing and reasoning under uncertainty. While useful in some areas, PGM suffers from its construction phase. This is known to be an NP-hard problem that can limit applications to some extent, especially in real-world applications. An ontology provides a structured body of knowledge characterized by a richness of meaning. In this white paper, we propose to use ontology's representational capabilities to enhance the process of creating PGMs. Of particular interest are Object Oriented Bayesian Networks (OOBN), which are extensions of standard Bayesian Networks (BN) using the object paradigm. We show how the semantical richness of on- tologies might be a potential solution to ad- dress the challenging field of structural learn- ing of OOBNs while minimizing experts in- volvement which is not always obvious to ob- tain. More precisely, we propose to set up a set of mapping rules allowing us to gener- ate a prior OOBN structure by morphing an ontology related to the problem under study to be used as a starting point to the global OOBN building algorithm