Meeting the Computational Demands of Nuclear Medical Imaging Using Commodity Clusters

• Authors: Wolfgang Karl;Martin Schulz;Martin Völk;Sibylle Ziegler
• Venue: ICCS '01 Proceedings of the International Conference on Computational Science-Part II
• Year: 2001

Even though Positron Emission Tomography (PET) is a relatively young technique within Nuclear Medical Imaging, it has already reached a high level of acceptance. However, in order to fully exploit its capabilities, computational intensive transformations have to be applied to the raw data acquired from the scanners in order to reach a satisfying image quality. One way to provide the required computational power in a cost-effective and efficient way, is to use parallel processing based on commodity clusters.These architectures are traditionally programmed using message passing. This, however, leads to a low-level style of programming not suited for the general user. In this work, a new programming environment based on a graphical representation of the application's behavior has been successfully deployed. The result is an image transformation application, which is both easy to program and fulfills the computational demands of this challenging application field.