Extended coronary trees by neural network based fusion of DSA-sequences and ECT-images

In order to support the clinical routine by integrated diagnoses, an existing coronary angiogram (DCI-System), computed tomography (CT) and emission computed tomography (ECT) imaging system is extended. To perform integrated imaging repetitively, standards and processing facilities are defined for each individual imaging system. Angiogram sequences are prestructured and viewed as typed dynamic data sets, allowing specific items to restrict regions of interest in an object-oriented and model-based way. Scintigraphic data are treated as rigid bodies in a CAD-like way. Both image modalities are described by labeled graphs. The connectivity between modality graphs is guided by a `rubber band´ approach performing, with an appropriate metric and energy function, an elastic matching. Within this hybrid structured approach, the advantages of neural networks for individual processing steps are demonstrated