Interactive Tomographic Reconstruction for Freehand SPECT using the GPU

Thesis by: Alexandru Duliu
Advisor: Nassir Navab
Supervision by: Tobias Lasser

Abstract

In modern medical practice, tomographic reconstructions offer a remarkable diagnostic tool. It started as a complement to already existing imaging techniques (like X-ray imaging) but soon spawned practices and therapies for which it has become invaluable. However the computational requirements associated with any kind of tomography are immense and can only be satised by dedicated workstations. In this work we adapt existing algorithms used in emission tomography for many-core GPU architectures. Using the graphics API of GPUs to speedup general computations is a popular technique nowadays. We however believe that too much of the computational resources of the GPU are wasted this way and that a more general approach to using the GPU'S necessary. To this end we used nVidia's new compute unied device architecture (CUDA) for all our computations. It offers us the ability to run C code natively on the GPU, avoiding the whole graphics related functionality altogether. With CUDA as a basis, we present a exible architecture for tomographic reconstructions that runs several times faster than on the CPU while removing all associated memory limitations. We then adapt our implementation for on-the-fly reconstructions for the emerging freehand Single Photon Emission Computed Tomography technology.

Resources

Literature

• C. Bischof and C. Van Loan: The wy representation for products of householder
• Doessel: Bildgebende Verfahren in der Medizin.
• Ntziachristos et al.: Looking and listening to light: the evolution of whole-body photonic imaging.
• Andrew Kerr, Dan Campbell, and Mark Richards: Qr decomposition on gpus.
• P.C. Lauterbur: Image formation by induced local interactions: examples employing nuclear magnetic resonance.
• NVIDIA: NVIDIA CUDA Compute Unied Device Architecture, June 2008.
• W.H. Press, S.A. Teukolsky, W.T. Vetterling, and B.P. Flannery: Recipes in C: The Art of Scientic Computing.
• Thomas Schiwietz: Acceleration of Medical Imaging Algorithms Using Programmable Graphics Hardware
• M. Slaney and AC Kak.: Principles of computerized tomographic imaging
• Steven W. Smith: The Scientist and Engineer's Guide to Digital Signal Processing