Dempster Shafer Approach for High Level Data Fusion Applied to the Assessment of Myocardial Viability

A modular data fusion system based on the Dempster-Shafer framework is presented. This system allows the building of any architecture of fusion by chaining elementary modules. Two types of modules are used, the numerical to symbolic conversion modules and the combination modules using logical rules of combination. The uncertainty is modeled by a basic belief assignment or the plausibility of each hypothesis. We applied our system to assess the Left Ventricular (LV) myocardial viability. The parameters taken into account are the LV contractile function extracted from tagged Magnetic Resonance Images (MRI) and the glucose metabolism rate obtained by Positron Emission Tomography (PET) imaging. The variables of interest are defined by the medical experts. The results are displayed on polar maps to give a geometrical information of the potential lesions.