Total variation regularization for edge preserving 3D SPECT imaging in high performance computing environments

Abstract

Clinical diagnosis environments often require the availability of processed data in real-time, unfortunately, reconstruction times are prohibitive on conventional computers, neither the adoption of expensive parallel computers seems to be a viable solution. Here, we focus on development of mathematical software on high performance architectures for Total Variation based regularization reconstruction of 3D SPECT images. The software exploits the low-cost of Beowulf parallel architectures. © Springer-Verlag Berlin Heidelberg 2002.